The New Hazardous Jobs and Worker Reallocation - OECD

Organisation for Economic Co-operation and Development

DELSA/ELSA/WD/SEM(2020)12

Unclassified English text only

24 July 2020

DIRECTORATE FOR EMPLOYMENT, LABOUR AND SOCIAL AFFAIRS EMPLOYMENT, LABOUR AND SOCIAL AFFAIRS COMMITTEE

The New Hazardous Jobs and Worker Reallocation

OECD SOCIAL, EMPLOYMENT AND MIGRATION WORKING PAPERS No. 246

JEL Codes: J28, J23, J81.

Keywords: Working conditions, workers’ reallocation, COVID-19 pandemic.

By Gaetano Basso (Bank of Italy, Italy), Tito Boeri (Bocconi University, CEPR and IZA, Italy & France), Alessandro Caiumi (Bocconi University, Italy) and Marco Paccagnella (OECD). This working paper has been an input to the analysis in Chapter 1 of the OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis on “COVID-19: From a Health to a Jobs Crisis”. Authorised for publication by Stefano Scarpetta, Director, Directorate for Employment, Labour and Social Affairs. All Social, Employment and Migration Working Papers are now available through the OECD website at www.oecd.org/els/workingpapers.

Gaetano Basso, Bank of Italy, [email protected] Tito Boeri, Bocconi University, CEPR and IZA, [email protected] Alessandro Caiumi, Bocconi University, [email protected] Marco Paccagnella, OECD Directorate for Education and Skills, [email protected]

JT03464237 OFDE

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the

delimitation of international frontiers and boundaries and to the name of any territory, city or area.

mailto:[email protected]




2 DELSA/ELSA/WD/SEM(2020)12

THE NEW HAZARDOUS JOBS AND WORKER REALLOCATION Unclassified

OECD Social, Employment and Migration Working Papers

www.oecd.org/els/workingpapers

OECD Working Papers should not be reported as representing the official views of the OECD or of its

member countries. The opinions expressed and arguments employed are those of the author(s).

Working Papers describe preliminary results or research in progress by the author(s) and are published to

stimulate discussion on a broad range of issues on which the OECD works. Comments on Working Papers

are welcomed, and may be sent to [email protected].

This series is designed to make available to a wider readership selected labour market, social policy and

migration studies prepared for use within the OECD. Authorship is usually collective, but principal writers

are named. The papers are generally available only in their original language – English or French – with a

summary in the other.

This document and any map included herein are without prejudice to the status of or sovereignty over any

territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city

or area.

Note by Turkey:

The information in this document with reference to “Cyprus” relates to the southern part of the Island. There

is no single authority representing both Turkish and Greek Cypriot people on the Island. Turkey recognises

the Turkish Republic of Northern Cyprus (TRNC). Until a lasting and equitable solution is found within the

context of the United Nations, Turkey shall preserve its position concerning the “Cyprus issue”.

Note by all the European Union Member States of the OECD and the European Union:

The Republic of Cyprus is recognised by all members of the United Nations with the exception of Turkey.

The information in this document relates to the area under the effective control of the Government of the

Republic of Cyprus.

© OECD 2020

You can copy, download or print OECD content for your own use, and you can include excerpts from

OECD publications, databases and multimedia products in your own documents, presentations, blogs,

websites and teaching materials, provided that suitable acknowledgment of OECD as source and copyright

owner is given. All requests for commercial use and translation rights should be submitted to

[email protected].

http://www.oecd.org/els/workingpapers



DELSA/ELSA/WD/SEM(2020)12 3


Acknowledgements

We thank Andrea Garnero, Eliana Viviano, seminar participants at the OECD and CEPR, and participants

to the annual VisitINPS conference for insightful comments. Francesco Armillei provided outstanding

research assistance.

The views expressed are those of the authors and do not necessarily reflect those of the Bank of Italy, nor

of the Eurosystem.



Abstract

This paper analyses several dimensions of workers’ safety that are relevant in the context of a pandemic.

We provide a classification of occupations according to the risk of contagion: by considering a wider range

of job characteristics and a more nuanced assessment of infection risk, we expand on the previous

literature that almost exclusively looked at feasibility of working from home. We apply our classification to

the United States and to European countries and we find that roughly 50% of jobs in our sample can be

considered safe, although a large cross-country variation exists, notably in the potential incidence of

remote working. We find that the most economically vulnerable workers (low-educated, low-wage workers,

immigrants, workers on temporary contracts, and part-timers) are over-represented in unsafe jobs, notably

in non-essential activities. We assess the nature of the reallocation of workers from unsafe to safe jobs

that is likely to take place in the years to come, and the policies that could mitigate the social cost of this

reallocation.

Résumé

Cet article analyse plusieurs dimensions relatives à la sécurité des travailleurs qui sont pertinentes dans

le contexte d'une pandémie. Nous proposons une classification des professions en fonction du risque de

contagion : en considérant un plus large éventail de caractéristiques professionnelles et une évaluation

plus nuancée du risque d'infection, nous développons la littérature précédente qui examinait presque

exclusivement la faisabilité du télétravail. Nous appliquons notre classification aux États-Unis et aux pays

européens et nous constatons qu'environ 50% des emplois de notre échantillon peuvent être considérés

comme non dangereux, bien qu'il existe une grande variation d'un pays à l'autre, notamment en ce qui

concerne l'incidence potentielle du travail à distance. Nous constatons que les travailleurs les plus

vulnérables sur le plan économique (travailleurs à faible niveau d’éducation, faiblement rémunérés,

immigrants, travailleurs sous contrat temporaire et travailleurs à temps partiel) sont surreprésentés dans

les emplois dangereux, notamment dans les activités non essentielles. Nous discutons la nature de la

réaffectation des travailleurs des emplois dangereux vers des emplois sûrs qui est susceptible de se

produire dans les années à venir, et les politiques qui pourraient atténuer le coût social de cette

réaffectation.



Table of contents

OECD Social, Employment and Migration Working Papers 2

Acknowledgements 3

Abstract 4

Résumé 4

1 Introduction 7

2 Literature Review 9

3 Methodology 11

4 Data and Results 16

Heterogeneity across sectors, occupations and firms 17

Heterogeneity across rural vs. urban areas 20

Heterogeneity across workers 25

Hazardous jobs and economic vulnerability 30

Targeting and reallocation 36

Essential and non-essential jobs: How risky? Who does them? 38

A risk-premium in unsafe jobs? 41

Workers training and reallocation 44

5 Concluding Remarks 46

References 48

Annex A. Statistical Annex 51

Annex B. Additional Figures 66

Tables

Table 1. Wage premia on safe jobs – United States 2018 42 Table 2. Wage premia on safe jobs – PIAAC data 43



Figures

Figure 1. Occupation classification: safe and unsafe jobs 14 Figure 2. Overall shares of jobs by category and country 17 Figure 3. Concentration indexes of job categories by economic sector 18 Figure 4. Concentration indexes of category 3 jobs by ISCO-2d occupation 19 Figure 5. Concentration indexes of job categories by metropolitan and non- metropolitan area 21 Figure 6. Shares of jobs by category and area 22 Figure 7. Concentration indexes of job categories by city, town and rural areas 23 Figure 8. Population shares across areas 24 Figure 9. Concentration indexes of job categories by firm size 25 Figure 10. Concentration indexes of job categories by age group 26 Figure 11. Overall shares of jobs by category and age 28 Figure 12. Concentration indexes of job categories by gender 29 Figure 13. Concentration indexes of job categories by education 30 Figure 14. Concentration indexes of job categories by income quintiles 31 Figure 15. Concentration indexes of job categories by household size 32 Figure 16. Concentration indexes of job categories by type of contract 33 Figure 17. Concentration indexes of job categories by working time arrangement 34 Figure 18. Concentration indexes of job categories by type of self-employment in the US 35 Figure 19. Concentration indexes of job categories by tenure 36 Figure 20. Concentration indexes of job categories in essential occupations 38 Figure 21. Concentration indexes of job categories by nativity status 39 Figure 22. Concentration indexes of workers’ characteristics in unsafe non-essential occupations 40



1. Epidemiological risk is an important dimension of worker safety, which recently gained enormous

importance due to the outbreak of the COVID-19 pandemic. Such risk is not evenly distributed across jobs

and workers. In this paper, we provide a method to classify occupations according to the risk of contagion

by analysing several dimensions of workers safety. We expand on the previous literature that almost

exclusively looks at the feasibility of working from home by analysing other dimensions of risk. Starting

from O*NET data, which describe the characteristics of jobs at the current level of technology, we create

four categories of jobs by gradually relaxing some safety constraints. This gives a more nuanced

characterisation of the epidemiological risk of occupations, encompassing working from home and various

degrees of physical proximity required to work.

2. We draw on data from the US Current Population Survey (CPS) and on harmonised labour force

survey data for European countries (EU LFS) to estimate the share of safe jobs in 27 countries. By using

the information contained in the O*NET survey, and building on the papers by Dingel and Neiman (2020[1])

and Boeri, Caiumi and Paccagnella (2020[2]), we classify jobs in four categories, characterised by

decreasing degree of safeness. To simplify the exposition, we label jobs belonging in the first three

categories as “safe jobs”, and those belonging in the last (residual) category as “unsafe jobs”. Each job

falls into a category if the average response of workers to an O*NET item is above (or below, depending

on how the question is formulated) a pre-determined threshold, starting from the safest type, being able to

work from home. Roughly 50% of jobs can be considered safe, although large cross-country variation

exists, notably in the share of jobs that can be done from home (which we consider to be the safest possible

form of work). We then take a detailed look at the characteristics of workers holding safe and unsafe jobs.

We find that the most economically vulnerable workers, i.e. low-educated workers, immigrants, low-wage

workers, workers on temporary contracts, and part-timers, are over-represented in unsafe jobs. Some of

these results are in line with previous studies, but we offer an analysis on an unprecedented set of worker

characteristics. Moreover, we show that about 60 percent of unsafe jobs are in non-essential occupations:

firms restructuring in these sectors may lead to a dramatic drop in labour demand hitting these

twice-vulnerable workers.

3. Another element that we consider is the potential for workers’ reallocation, which will be a crucial

issue for labour markets in a scenario where the risk of a pandemic remains significant. In a world subject

to the risk of recurrent pandemic waves, we envisage two broad forms of reallocation: the first is from

unsafe to safe occupations, and the second is towards occupations that, despite being unsafe, are

“essential”, and thus not subject to lockdown measures. A better match between workers and jobs along

the infection risk dimension should be pursued, by ensuring for instance that workers more prone to incur

severe forms of the disease move away from unsafe jobs. For this reason, we also provide preliminary

indications on the distribution of safe and unsafe jobs across sectors and occupations, as well as dedicating

special attention to those characteristics of workers, such as age, that are supposedly related to higher

mortality risks from COVID-19. On top of this, changes in demand for certain goods or services are also

likely to cause changes in the occupational structure, and a reallocation of workers across different sectors.

For this reason, we conclude the paper with some broad considerations on social protection and the need

for policies that foster such reallocation.

1 Introduction



4. Our contribution to the literature on the labour market consequences of the COVID-19 pandemic,

which we survey in detail in Section II, is threefold. First, we consider a broader definition of hazardous

and non-hazardous jobs: we focus not only on jobs that can be performed remotely, but we include also

jobs that, while requiring presence on the workplace, comply with physical distancing protocols as they

involve only infrequent contacts with other persons. Second, rather than examining a single country, our

paper carries out a systematic analysis of the labour market of 27 countries. Third, we characterise the

heterogeneity in safe jobs across a large number of jobs and firm characteristics and we discuss the

potential implications in terms of workers’ reallocation and workers-to-jobs matching along the

epidemiological risk dimension.

The remainder of the paper is structured as follows. Following the literature review in section II, in Section

III we present our methodology to classify jobs according to their level of infection risk. In Section IV we

apply our classification to 27 countries. After estimating the incidence of safe jobs, we analyse the

heterogeneity according to a large number of jobs and workers characteristics. Section V concludes by

discussing the implications of our results for the reallocation of workers that will have to take place in a

world characterised by a significant pandemic risk.



5. Since the first appearance of the pandemic, the literature on the economic consequences of

COVID-19 has been growing rapidly. A number of studies have attempted to estimate the share of jobs

that can be performed at home or remotely, and which are therefore compatible with the most restrictive

lockdown measures put in place to contain the pandemic. The earliest papers are Dingel and Neiman

(2020[1]), who focus on the US and rely on the O*NET surveys to estimate the feasibility of working from

home, and Boeri, Caiumi and Paccagnella (2020[2]), who expand the perspective of the analysis by taking

into account also jobs that entail only sporadic or no personal interactions. They complement information

from O*NET with personal judgements and classify jobs in six EU countries.

6. Relying on American Time Use Survey (ATUS) data, Hensvik, Le Barbanchon and Rathelot

(2020[3]) compute the share of hours worked at home for American workers by occupation and industry.

They conclude that 15% of the overall working hours are carried out from home and that workers in high-

skilled occupations work more hours at home than workers in less skilled occupations. Mongey, Pilossoph

and Weinberg (2020[4]) combine the two approaches by using ATUS data to empirically validate the

information on the feasibility of working from home (derived from O*NET) with the actual habits of workers.

Linking such indicators with information from the Bureau of Labor Statistics Current Population Survey

(BLS CPS) and Panel Study of Income Dynamics (PSID), they conclude that workers in occupations that

cannot be done remotely are more likely to be economically vulnerable (i.e. they are less likely to have a

college degree, to have health insurance, to have liquid assets, to be white, to be US natives). They also

find that metropolitan areas (MSA) with lower shares of employment in work-from-home jobs before

COVID-19 experienced smaller declines in mobility, as measured using cell phone data, and that both

occupations and types of workers predicted to be employed in low work-from-home jobs experienced

greater declines in employment with respect to pre-pandemic months.

7. Similar exercises have been carried out for a larger set of countries. Saltiel (2020[5]) focuses on

ten low- and middle-income countries, using data from the World Bank Skills Toward Employability and

Productivity (STEP) survey (as information from O*NET surveys may not be representative of the task

content of occupations in developing countries). He concludes that only 13% of workers can work from

home in the set of ten countries analysed, and that the feasibility of home working is positively correlated

with high-paying occupations, educational attainment, formal employment status and household wealth.

8. Gottlieb, Grobovšek and Poschke (2020[6]) analyse how the share of employment that can work

from home varies with the country income level. They build a micro level dataset for 57 countries and find

that the share of workers in urban areas that can work from home is about 20% for poor countries, while it

is about twice as high in rich countries. This is largely due to the higher share of self-employed workers

not able to work remotely in poor countries.

9. Barbieri, Basso and Scicchitano (2020[7]) use the Sample Survey of Professions, the Italian

equivalent to O*NET, to study risk profiles for workers in the Italian labour market. They find that workers

in occupations that are exposed to infection and disease risks tend to work in close physical proximity to

other people, and that several other sectors, mainly related to personal services, leisure and recreation

are not directly exposed to infections and diseases, but need physical proximity to operate. They also find

that groups who are at risk of contagion and complications from COVID-19 (mainly male, and workers

aged above 50) work in sectors that are either little exposed to physical proximity (such as agriculture),

2 Literature Review



that were under lockdown or can potentially work remotely (such as public administration and some

education subsectors). Different from this paper, however, Barbieri et al. build only relative measures of

proximity risks and disease exposure and do not quantify the absolute number of workers at risk.

10. Adopting a different perspective, Adams-Prassl et al. (2020[8]) study the within-occupation

heterogeneity in the feasibility of working from home and reach different conclusions from other studies.

Using large surveys from the US and UK, they find that the share of tasks that can be done from home is

not constant across workers within occupations or industries: their ability to work from home varies both

across and within occupations and sectors.

11. In an early attempt to associate workers’ safety risk and actual contagion, Lewandowski (2020[9])

estimates country-specific levels of COVID-19 spread (from the Johns Hopkins CSSE) and social contacts

and diseases exposure indicators (built from O*NET and the European Working Condition Survey data).

Analysing 26 EU countries, he finds that higher levels of occupational exposure to contagion are positively

correlated with faster growths in COVID-19 cases and deaths, in particular for workers aged 45-64.

12. Other recent works analyse the heterogeneity in the impact of COVID-19. Yasenov (2020[10])

provides evidence of the distributional effects of stay-at-home orders caused by the pandemic in the United

States. He links O*NET data to the American Community Survey (ACS) to understand the types of workers

holding jobs that can be done from home, finding that workers with lower wages, lower levels of education,

younger workers, ethnic minorities and immigrants are less likely to work in occupations that are suitable

for remote working. Using data from the Survey of Business Uncertainty (SBU), Barrero, Bloom and Davis

(2020[11]) quantify the short-term reallocative impact of the pandemic: they estimate that COVID-19 shock

caused 3 new hires for every 10 layoffs. According to the authors much of this impact will persist, with 42%

of recent layoffs that will become permanent job losses. They also develop forward-looking excess job

reallocation measures (the difference between job turnover and net employment growth), which rise

sharply after the beginning of the pandemic.

13. Finally, a set of papers focused on the unequal effects of the pandemic on different group of

workers. Borjas and Cassidy (2020[12]) rely on CPS data to find that the negative employment shock caused

by the pandemic in the United States hit immigrants severely. They find that this can be explained by the

fact that the rate of job loss rose for immigrants compared to natives (partly because immigrants are less

likely to work in occupations that can be performed remotely) and that job-finding rate for unemployed

immigrants fell compared to natives. OECD (2020[13]) also documents that in the sectors most affected by

the COVID-19 containment measures, non-standard forms of employment are over-represented.

14. Fasani and Mazza (2020[14]) focus on immigrants in the European Union. They find that in EU27

13% of immigrants are employed in “essential” occupations that were not affected by the most restrictive

lockdown measures. For instance, they account for 20 percent or more of employment in occupations like

cleaners and helpers, labourers in mining and construction, personal care workers, and food processors.

15. Using real time survey data from the project REPEAT for a number of OECD countries, Foucault

and Galasso (2020[15]) documents unequal lockdown effects across categories of workers. Low-educated

workers, blue collars and low-income service workers were more likely to have suspended working

activities and low-educated workers less likely to work from home. Adjustments took place over the

lockdown weeks, with higher shares of workers becoming active both from home and from their workplace,

but such adjustments benefitted mostly highly educated workers and white collars.



16. Within the context of the COVID-19 pandemic, safe jobs are those that can be carried out with a

minimal risk of being infected and of spreading the virus. It is important to clarify upfront that this

classification is necessarily based on the way jobs were carried out in “normal times”, i.e. before the

outbreak of the COVID-19 pandemic.

17. This exercise is therefore informative as to the number of jobs that will likely not require any major

organisational change. It is certainly possible that jobs that we are now classifying as unsafe will be

reorganised in ways that allow them to be performed while minimising the risk of contagion, although we

are not yet in a position to assess the effects of these changes on productivity. A good example is primary

school teachers: we classify them as “unsafe”, because O*NET data tell us that performing this job involve

daily physical contacts with a large number of people, and still many teachers have managed to switch to

online lectures in many of the countries that have implemented prolonged lockdown measures.

18. We classify jobs in four categories: three are based on different definitions of safety and the last

is built as a residual category, containing jobs that can be deemed “unsafe” (again, under norma l

circumstances).

19. The first category contains all occupations that can be potentially performed remotely (category 1

jobs). These jobs do not require workers to leave their home, nor to interact with co-workers or customers

in person. This measure provides the most restrictive definition of safety, as it essentially reduces the risk

of work-related contagion to zero.

20. The second category relaxes slightly these constraints by adding jobs that require at most a low

level of physical proximity on the workplace and a limited number of interactions with external customers

and the public. Arguably, these jobs do not pose significant risks to workers’ health under a pandemic.

21. The third category still requires a low level of physical proximity, but allows for the inclusion of jobs

that involve a higher degree of interactions with external customers. The need to interact with external

customers potentially increases the size of the network the worker is exposed to, which is clearly an

important element to consider in the context of a pandemic. These jobs are likely characterised by an

element of “mobility”, either because the workers have to visit customers, or because customers have to

visit the workers.

22. Any job that cannot be done from home presents an additional risk factor related to commuting,

which mechanically increases the risk of infection by increasing human interaction. Unfortunately, the data

we use do not contain information on commuting habits (although we do look at some rough proxy such

as whether people live in urban or rural areas). This additional risk is therefore very difficult to quantify,

also because it varies significantly in ways that are difficult to predict with available data, such as whether

the worker use private or public transportation, how much time is spent in commuting, and whether it takes

place at peak or off-peak hours.

23. These first three categories are constructed so that category 1 is a subset of category 2, which is

in turn a subset of category 3. Category 3 therefore consists of all jobs that can be considered safe

according to our least restrictive definition of safety (and it includes category 1 and category 2 jobs). The

fourth and last category is instead a residual category that contains all remaining jobs, which we label

“unsafe” from now on as they entail a relatively high risk of being infected by COVID-19.

3 Methodology



24. In order to assign occupations to these categories, we rely on data from the Bureau of Labor

Statistics (BLS) O*NET survey, which contains information on the most distinctive traits of each job in the

United States. Building on the work of Dingel and Neiman (2020[1]), and on our previous classification of

occupations (Boeri, Caiumi and Paccagnella, 2020[2]), we first select 27 questions from the “Work context”

and “Work Activities” sections of O*NET database that, according to our judgment, provide information on

the feasibility of working from home. As in Dingel and Neiman (2020[1]), we classify a job as not doable

from home if the average response of workers to an item is above (or below, depending on how the

question is formulated) a pre-determined threshold. The answers to each item can take values ranging

from 1 to 5, where higher values denote a stronger intensity or higher frequency of the trait under scrutiny.

The list of the questions considered and the conditions imposed follow.

25. From the “Work context” section:

Q4 – "Average respondent says they use email less than once per week" (value < 3.0/5.0)

Q14 – "Average respondent says they deal with violent people at least once a week" (value >

4.0/5.0)

Q16 – "Average respondent says they work indoors, in an environment not controlled, almost every

day" (value > 4.5/5.0)

Q17 – "Average respondent says they work outdoors, exposed to all conditions, almost once per

week at least" (value > 3.5/5.0)

Q18 – "Average respondent says they work outdoors, under cover, almost every day" (value >

4.5/5.0)

Q19 – "Average respondent says they work in an open vehicle or operating equipment almost

every day" (value > 4.5/5.0)

Q20 – "Average respondent says they work in a closed vehicle or operate enclosed equipment

almost every day" (value 4.5/5.0)

Q23 – "Average respondent says they are exposed to extreme temperatures almost every day"

(value > 4.5/5.0)

Q29 – "Average respondent says they are exposed to diseases or infection at least once a month"

(value > 3.0/5.0)

Q30 – "Average respondent says they are exposed to high places at least once a week" (value >

4.0/5.0)

Q31 – "Average respondent says they are exposed to hazardous conditions at least once a week"

(value > 4.0/5.0)

Q32 – "Average respondent says they are exposed to hazardous equipment at least once a week"

(value > 4.0/5.0)

Q33 – "Average respondent says they are exposed to minor burns, cuts, bites, or stings at least

once a week" (value > 4.0/5.0)

Q34 – "Average respondent says they are sitting less than half the time" (value < 2.0/5.0)

Q36 – "Average respondent says they spend more than about half the time climbing ladders,

scaffolds, or poles" (value 3.5/5.0)

Q37 – "Average respondent says they spend more than about half the time walking or running"

(value 3.5/5.0)

Q43 – "Average respondent says they wear common protective or safety equipment more than

once per month" (value 3.5 > 5.0)

Q44 – "Average respondent says they wear specialised protective or safety equipment more than

once per month" (value 3.5 > 5.0)



26. From the “Work activities” section:

Q4 – "Inspecting equipment, structures or materials is important/very important" (value > 3.5/5.0)

Q16 – "Performing general physical activities is important/very important" (value > 3.5/5.0)

Q17 – "Handling and moving objects is important/very important” (value > 3.5/5.0)

Q18 – “Controlling machines and processes is very important" (value > 4.0/5.0)

Q19 – “Working with computers is not important" (value < 1.5/5.0)

Q20 – “Operating vehicles, mechanised devices or equipment is important/very important” (value

3.5/5.0)

Q22 – “Repairing and maintaining mechanical equipment is important/very important" (value >

3.5/5.0)

Q23 – “Repairing and maintaining electronic equipment is very important" (value 4.0/5.0)

Q29 – “Assisting and Caring for others is important/very important" (value 3.5/5.0)

27. If any such condition is true, we classify a job as not suitable for remote working. For instance, if

for a given job the average answer to question 4 from the “Work context” section (“How frequently does

your current job require electronic email?”) is lower than 3.0 (where 3.0 represents the option “once a

month or more but not every week”), we consider that job as not suitable for remote working, and we

therefore exclude it from category 1.

28. This procedure generates a dummy value for each job, which we then aggregate into 3-digit

occupational codes as follows. First, we map O*NET occupations to SOC occupations through simple

averages whenever the correspondence is not 1-to-1. Next, using as weights the 2018 US employment

shares of SOC occupations from the BLS Employment Projections program, we map values from SOC

occupations into ISCO 3-digit codes that identify occupations in the EU LFS data through weighted

averages.1 Thus, for each ISCO 3-digit code, we obtain a coefficient, ranging from 0 to 1, proxying the

share of jobs that can be carried out remotely according to our definition and to the description of

occupations contained in O*NET. For the US data, instead, we convert the occupational codes of the CPS

to SOC codes, and we link then our taxonomy directly at the SOC level, without walking through the ISCO

classification. One caveat to this approach is that we start from the characteristics of US jobs, namely

technology and labour market conditions, and we map these to European jobs. This exercise necessarily

entails some measurement error as long as technology differs across countries and European occupations

are carried out differently with respect to US ones. Thus, our results need to be interpreted as if the US

occupational technology was in place for each labour market analysed.

29. In order to identify the jobs belonging to category 2 and category 3 we use other questions from

O*NET, in particular those on physical proximity and contacts with public and customers:

“Work context”: Q21 – “How physically close to other people are you when you perform your current

job?” (value 3.5/5.0)

“Work activities”: Q32 – " How important is performing for or working directly with the public to the

performance of your current job?” (value 3.0/5.0)

1 While the objective of the Employment Projections (EP) program by the U.S. Bureau of Labor Statistics is to provide

estimates of occupational trends over a 10-year projection period, we use it as crosswalk as it is the most

comprehensive collection of occupation-level data in the US. For this purpose, we focus on the base-year only (2018

in our case) whose data are actual employment figures derived from the OES program, CES program, QCEW and

CPS. Indeed, the advantage of using EP program’s data is to cover the universe of US occupations hinging on a

combination of sources: nonfarm wage and salary employment is covered by OES, CES and QCEW, whereas

agricultural industry employment, self-employed workers, and workers in private households are covered by the CPS.



30. Category 2 includes, in addition to jobs that can be conducted remotely, those that entail low

physical proximity and limited exposure to customers and to the public. In other words, category 2 includes

all jobs already present in category 1, plus those that feature low average answers to both Q21 and Q32.2

Category 3 further relaxes this criterion by allowing jobs reporting higher values for Q32 (“Performing for

or working directly with public” can display values above 3.0), as such jobs may also require substantial

exposure to external persons, while still maintaining low physical proximity.

31. Being remote working the safest working arrangement, category 1 is, by construction, a subset of

category 2, which in turn is a subset of category 3. All the remaining occupations are classified as unsafe

(see Figure 1). Category 3 is thus the least restrictive measure of safety, as it includes workers who may

meet many different people during their working activity, even though without close personal contacts.

32. At the same time, we require all jobs in category 2 and 3 to meet the same condition imposed for

jobs belonging to category 1 with respect to question Q29 of the “Work section”: "Average respondent says

they are exposed to diseases or infection at least once a month". Any job exposed to diseases or infection

according to Q29 is classified as unsafe, regardless of whether the conditions on low physical proximity

and few contacts are met. Finally, with a procedure analogous to the one described before, we map dummy

values from O*NET occupations to SOC occupations and then to ISCO 3-digit codes for category 2 and

category 3 as well.

Figure 1. Occupation classification: safe and unsafe jobs

2 The 3.5 threshold for question Q21 corresponds to “Work more distant than arm’s length”, the 3.0 threshold for

question Q32 corresponds to “Performing for or working directly with public is important”.



Note: The figure describes the relationship among the job categories identified according to different definition of safety. Remote working is the

safest working arrangement so that category 1 (light grey) is, by construction, a subset of category 2 (grey), which in turn is a subset of category

3 (dark grey), as the latter relies on the broadest definition of safety. Occupations not included in these three categories are classified as unsafe

and belong to the residual category (black).

33. Having classified all occupations in one of these four categories, we can then compute the shares

of safe jobs (where we consider a job safe if it belongs in one of the first three categories) in countries for

which we have access to occupational identifiers from labour force surveys.

34. Some occupations well represent the categories that we identified.3 For category 1, for instance,

some ISCO codes report a coefficient equal to 1, the maximum value, indicating that these occupations

have the highest number of workers who can potentially work remotely. Examples of such occupations

are: information and communications technology service managers (ISCO 133), finance professionals

(241), legal professionals (261), sales and purchasing agents and brokers (332), and secretaries (412). By

construction coefficients for these ISCO codes will also be equal to 1 for category 2 and category 3.

35. For this reason, rather than looking at the highest coefficients for category 2, it is more interesting

to look at the ISCO codes that feature the largest increases in coefficients when moving from category 1

to category 2. By doing this, we identify occupations where many workers cannot work from home but

whose job entails low physical proximity and limited exposure to customers and the public.

36. The five occupations for which the coefficient increases the most when moving from category 1 to

category 2 are: mixed crop and animal producers (613), blacksmiths, toolmakers and related trades

workers (722), wood processing and papermaking plant operators (817), other stationary plant and

machine operators (818), and vehicle, window, laundry and other hand cleaning workers (912).

Conversely, the occupations for which the coefficient increases the most when moving from category 2 to

category 3 are social and religious professionals (263) and heavy truck and bus drivers (833). Wood

treaters, cabinet-makers and related trade workers (752), subsistence crop farmers (631) and

telecommunications and broadcasting technicians (352) also report lower but sizeable increases.

37. Finally, a few examples of “unsafe” occupations reporting a coefficient equal to 0 for category 3

(i.e. ISCO codes with the highest number of workers in jobs belonging to the residual category) are: medical

doctors (221), primary school and early childhood teachers (234), nursing and midwifery associate

professionals (322), food preparation assistants (941), and waiters and bartenders (513).

3 Detailed information on category coefficients for all ISCO 3-digit codes can be found in Annex A.



38. We apply our classification of occupations to labour force survey data for 2018 for EU countries

(harmonised EU LFS data) and for the United States (CPS, averaging the 12 monthly waves), ending up

with a sample of 27 countries: Austria, Belgium, Croatia, Cyprus4, the Czech Republic, Denmark, Estonia,

Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Luxembourg, the Netherlands,

Norway, Portugal, Romania, the Slovak Republic, Spain, Sweden, Switzerland, the United Kingdom and

the United States.5

39. For each country, Figure 2 reports the share of workers (either employees or self-employed

persons) holding jobs that can be considered “safe” according to our taxonomy (categories 1, 2 and 3).

Workers in safe jobs make up more than 50% of overall employment in 22 countries out of 27. Overall, the

average share of safe jobs, weighted by the number of employed individuals in each country, is 51.9%,

with Luxembourg reporting the highest share (60.7%) and Spain the lowest (44.1%). The black portion of

the bar represents the share of “unsafe” jobs for each country (residual category).

40. The figure also illustrates a breakdown according to the three categories defined above. The

variation across countries in the incidence of the different categories is quite large.

41. The share of workers in category 1 (i.e. jobs that can be performed from home), represented by

the first light grey portion of the bars, ranges from 17.0% in Romania to 43.5% in Luxembourg, with a

weighted average of 31.7% for the whole sample. As for category 2, given by the sum of the first two

portions of the bars, Spain reports the lowest share (37.6%) and Luxembourg the highest (53.1%), with an

overall weighted average of 44.9%. When we add also the jobs that require interaction with customers

(category 3), the cross-country variation declines. This implies that countries in which fewer jobs can be

done from home (category 1) have relatively more jobs that require infrequent personal contacts and that

belong to category 2 or 3.6

4 A. Note by Turkey: The information in this document with reference to “Cyprus” relates to the southern part of the

Island. There is no single authority representing both Turkish and Greek Cypriot people on the Island. Turkey

recognises the Turkish Republic of Northern Cyprus (TRNC). Until a lasting and equitable solution is found within the


B. Note by all the European Union Member States of the OECD and the European Union: The Republic of Cyprus is

recognised by all members of the United Nations with the exception of Turkey. The information in this document relates

to the area under the effective control of the Government of the Republic of Cyprus.

5 We could not exploit data from Bulgaria, Lithuania, Malta, Poland and Slovenia due to missing information on many

of the dimensions analysed.

6 In the US the fraction of jobs that require contacts with customers in addition to co-workers and that are not suitable

for remote arrangements, but are still considered safe according to our taxonomy, reaches 8%, the highest percentage

of our sample.

4 Data and Results



Figure 2. Overall shares of jobs by category and country

Note: The figure shows the percentage of workers holding a job belonging to the different categories of our taxonomy across the 27 countries

of the sample. Data refer to 2018.

Source: Current Population Survey (CPS) and European Labour Force Survey (EU LFS).

Heterogeneity across sectors, occupations and firms

42. Figure 3 displays the distribution of jobs in categories 1 and 3 across economic sectors, following

the NACE Rev. 2 (2008) classification, pooling data from all countries in the sample.7 Histograms display

concentration indexes, computed as the ratio between the share of jobs of category i in sector j over the

share of category i in total employment. A value greater (lower) than one of the index, denotes sectors

over-represented (under-represented) in that specific category.

43. In sectors like “Financial and insurance activities”, “Professional, Scientific and Technical activities”

and “Information and Communication”, the share of jobs that can be carried out remotely is twice as large

as in the overall labour market. In these sectors, the share of jobs that can be considered safe according

to our broader definition (category 3 jobs) is close to 100%. At the other extreme of the distribution, sectors

like “Human health and social work activities”, “Accommodation and food service activities” and

“Households as employers” report the lowest concentration indices for category 3 jobs. The agricultural

sector represents a peculiar case: it has the second lowest share of jobs that can be conducted remotely,

but when it comes to the share of safe jobs the concentration index is above 1. This can be explained by

7 We take a weighted average of the country-specific employment shares taking as weights the relative size of

employment in each country and category. Information at the country level can be found in Annex A.



the fact that many activities in agriculture can be carried out preserving physical distance among workers

and only sporadic interactions are needed mostly with co-workers.

Figure 3. Concentration indexes of job categories by economic sector

Note: The figure shows concentration indexes for category 1 and category 3 jobs by economic sector (Nace rev 2). Concentration indexes are

computed as the ratio between the share of jobs of category i in sector j over the share of category i in total employment, pooling data from the

27 countries of the sample. Numbers greater (lower) than one (vertical dashed bar) denote over-representation (under-representation) in that

specific category. Data refer to 2018.


44. Figure 4 displays the five safest occupations and the five least safe occupations at ISCO 2-digits.

In occupations like health, personal care, food preparation, and refuse workers, the share of unsafe jobs

is particularly large. Combined with Figure 3, the large cross-occupation dispersion in the incidence of safe

jobs is an indicator of the likely extent of job and worker reallocation that will follow the pandemic. Indeed,

given that job reallocation is the sum of employment-weighted firm-level employment growth rates, it is

increasing in the heterogeneity of the impacts of the pandemic.

45. There is evidence that firms and workers are redirecting search away from these risky occupations

(Hensvik, Le Barbanchon and Rathelot, 2020[3]), and surveys of employers indicate that firms implementing

physical distancing may suffer marked declines in productivity and even in capacity and production levels.

Thus, both labour demand and supply factors may induce substantial reallocation of jobs and workers

away from these relatively unsafe occupations and sectors.

46. A significant downsizing can be expected in firms operating in sectors with lower productivity and

in which labour supply will decline as a consequence of COVID-19. Before the pandemic, such firms were



employing a significant portion of the overall workforce, but labour supply and labour demand are now self-

reinforcing in reducing employment. Sectors like “Arts, entertainment and recreation” and “Accommodation

and food service activities” jointly employ 29.1 million workers in our sample, corresponding to slightly less

than 8% of total employment, a share comparable to half of that of the public sector.8

Figure 4. Concentration indexes of category 3 jobs by ISCO-2d occupation

Note: The figure reports the 5 occupations at the ISCO 2-digit level at the top and at the bottom of the distribution of concentration indexes for

category 3. Concentration indexes are computed as the ratio between the share of jobs of category 3 in occupation j over the share of category

3 in total employment, pooling data from the 27 countries of the sample. ISCO code 95 has been dropped due to inconsistencies between ONET

and ICP INAPP data. Numbers greater (lower) than one (vertical dashed bar) denote over-representation (under-representation) in that specific

category. Data refer to 2018.


47. Job creation will likely be stronger in the health sector and in those industries that serve the health

sector also as a result of stronger public expenditures in these strategic activities. Such industries revolve

around two main poles: the pharmaceutical industry and healthcare services. The first also encompasses

the chemical industry, part of the packaging industry, research centres, logistics and pharmaceutical

wholesale and retail trade. The second encompasses the supply of goods (manufacturing of hospital

8 Construction also features a large share of unsafe jobs. However, injury risk was particularly high in this sector even

before the pandemic, and not all countries enforced the lockdown in construction. Thus, we consider it to be a rather

special case.



equipment and devices) and services (cleaners, hospital assistance and security, linen rental, food

services), the management of special waste and the insurance sector.9

Heterogeneity across rural vs. urban areas

48. Another important dimension of potential reallocation relates to the place of residence of

individuals. Higher population density is likely to increase the risk of infection, irrespective of the

characteristics of jobs, as workers are for instance more likely to commute by public transportation. On the

other hand, it is also possible that metropolitan and non-metropolitan areas have different occupational

structures, and thus a different prevalence of safe or unsafe occupations. Commuting is not captured in

our data, but we can nevertheless look at the incidence of safe occupations according to the population

density in the place of residence.10

49. Figure 5 documents that metropolitan areas feature a higher share of jobs that can be done from

home (category 1 jobs) compared to the whole economy, whereas safe jobs in general (including those in

category 3) seem to be more evenly distributed across areas.11 Vertical bars measure the cross-country

variation in these concentration indexes.

50. In Figure 6 we show the additional share of jobs that each category brings to the total share of

safe jobs as we gradually expand the definition of safety to go beyond remote working. We also plot the

share of “unsafe” jobs, so that the total of each pie adds up to total employment. In line with Figure 5, the

share of safe jobs is similar in urban and rural areas, but non-metropolitan areas show a higher share of

jobs belonging to category 2 but not to category 1 (i.e. jobs that entail low physical proximity and limited

exposure to customers and public, but that cannot be conducted remotely). Likely, part of such jobs

belongs to the agricultural sector, mostly present in rural and scarcely populated areas.

51. Figure 7 provides a more granular description of areas, relying on the EU LFS methodology to

describe different degrees of urbanisation (unfortunately, this breakdown is not available for the US).12

According to Figure 7, jobs that can be performed remotely are over-represented in cities and under-

represented in rural areas, whereas unsafe jobs are slightly over-represented in rural areas.

52. Figure 8 shows no stark differences in the percentage of population living in different areas by age

group. People aged 25 to 34 years are over-represented in cities, whereas the age distribution is more

skewed towards the oldest individuals in rural areas. Yet the demand for essential services by the elderly

population in domains like health and personal care (where unsafe jobs are prevalent; see Figure 4), might

contribute to explain the prevalence of unsafe jobs in rural areas.

9 Estimates on the size and features of these industries are not possible with our data, which report only aggregate

information in terms of economic sectors.

10 Clearly the availability and usage of public transportation and the commuting habits of workers more in general are

important dimensions that affect the risk of contagion.

11 Metropolitan areas are defined as areas with more than 100,000 individuals.

12 Cities are defined as areas where at least 50% of residents live in high-density clusters, i.e. contiguous grid cells of

1 km2 with a density of at least 1,500 inhabitants per km2 and a minimum population of 50,000. In rural areas, more

than 50% of the population lives in rural grid cells, defined as grid cell outside high-density clusters and urban clusters

(cluster of contiguous grid cells of 1 km2 with a density of at least 300 inhabitants per km2 and a minimum population

of 5,000). Finally, towns are defined as areas where less than 50% of the population lives in rural grid cells and less

than 50% live in high-density clusters.



Figure 5. Concentration indexes of job categories by metropolitan and non- metropolitan area

Note: The figure shows concentration indexes of job categories by living area. Non-metropolitan areas are those with less than 100,000

inhabitants. For European countries, Towns and suburbs and Rural areas (as defined in EU LFS) are aggregated into Non-metropolitan.

Concentration indexes are computed as the ratio between the share of jobs of category i in area j over the share of category i in total employment,

pooling data from the 27 countries of the sample. Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-

representation) in that specific category. Vertical bars measure one standard deviation above and below the cross-country average of

concentration indexes. Data refer to 2018.


53. Unsafe jobs are also over-represented in small firms (Figure 9). This is consistent with the risk

exposure in the accommodation and food sectors, but may also have to do with the presence of fixed costs

in organising remote working. Early evidence from real-time privately-owned data in the United States also

points to a concentration of job losses in small firms (Chetty et al., 2020[16]). In Europe, small and medium

enterprises (SMEs) have been among the main intended beneficiaries of the short-time work schemes

introduced to mitigate the impact on unemployment of the crisis. Indeed, STW schemes have been

extended in several countries to cover also the small business sector (Giupponi and Landais, 2020[17]),

although the take-up rate among SMEs has been relatively low (OECD, 2020[18]).



Figure 6. Shares of jobs by category and area

Note: The figure shows the percentage of workers with a job belonging to the different categories of our taxonomy across the 27 countries of

our sample, by area of residence. For European countries, Towns and suburbs and Rural areas are aggregated into Non-metropolitan. Data

refer to 2018.




Figure 7. Concentration indexes of job categories by city, town and rural areas

Note: The figure shows concentration indexes of job categories by area of residence. Definition of areas follows the methodology adopted by

Eurostat. In cities, at least 50% lives in high-density clusters, defined as contiguous grid cells of 1 km2 with a density of at least 1,500 inhabitants

per km2 and a minimum population of 50,000. In rural areas: more than 50% of the population lives in rural grid cells, defined as grid cell outside

high-density clusters and urban clusters (cluster of contiguous grid cells of 1 km2 with a density of at least 300 inhabitants per km2 and a

minimum population of 5,000). In towns, less than 50% of the population lives in rural grid cells and less than 50% live in high-density clusters.

Concentration indexes are computed as the ratio between the share of jobs of category i in living area j over the share of category i in total

employment, pooling data from the 26 countries of the sample (no data available for the United States). Numbers greater (lower) than one

(horizontal dashed bar) denote over-representation (under-representation) in that specific category. Vertical bars measure one standard

deviation above and below the cross-country average of concentration indexes. Data refer to 2018.

Source: European Labour Force Survey (EU LFS).



Figure 8. Population shares across areas

Note: The figure shows the percentage of the overall population living in each area, by age group. Pooled data from 26 countries of our sample

(data are not available for the US). Data refer to 2018.




Figure 9. Concentration indexes of job categories by firm size

Note: The figure shows concentration indexes for job category by size of the firm. Concentration indexes are computed as the ratio between the

share of jobs of category i in size group j over the share of category i in total employment, pooling data from 26 countries of the sample (data

are not available for the US). Numbers greater (lower) than one (vertical dashed bar) denote over-representation (under-representation) in that

specific category. Data refer to 2018.


Heterogeneity across workers

54. The results presented so far suggest that a large fraction of workers (about one sixth of the total)

may be at risk of being dismissed or having to face significant changes in the way they work, should the

pandemic be long lasting. The high prevalence of unsafe jobs in small firms might also results in the exit

of business units and firms closures: job losses driven by firm exit, as opposed to downsizing of continuing

units, would be less gradual. Traditional measures to smooth over time labour market adjustments – such

as employment protection, and short-time work – or even the banning altogether of layoffs introduced in

some countries, may prove rather ineffective in this context and are becoming unsustainable with the

prolongation of the crisis.

55. It is therefore of uttermost importance to evaluate which are the categories of workers likely to be

most involved in this reallocation away from the epidemiological hazard. In particular, a key issue is

whether these are the same workers already in a vulnerable position at the outset of the crisis.

56. The analysis of the distribution of safe and unsafe jobs by worker characteristics can also shed

light on the health complications potentially associated to a second wave of the pandemic. A key dimension

to be considered in this context is age. A consensus seems to be emerging in the medical literature that



males over the age of 60 are significantly more likely to develop serious or critical forms of COVID-19

infection (Poletti et al., 2020[19]). The evidence to date also points overwhelmingly to much higher mortality

rates for elderly men.13

Figure 10. Concentration indexes of job categories by age group

Note: The figure shows concentration indexes of job categories by age group. Age groups reported are the two at the extremes of the working-

age interval: 15-24 years and 55-65 years. Concentration indexes are computed as the ratio between the share of jobs of category i in group j

over the share of category i in total employment, pooling data from the 27 countries of the sample. Numbers greater (lower) than one (horizontal

dashed bar) denote over-representation (under-representation) in that specific category. Vertical bars measure one standard deviation above

and below the cross-country average of concentration indexes. Data refer to 2018.


57. The average age of workers in safe jobs is around 43, with little variation among the three different

categories.14 The average age of workers in jobs that we label unsafe is slightly lower, at around 41. This

is in line with Figure 10, which shows concentration indices for safe jobs (category 1 to category 3) at the

two extremes of the age distribution: for young (15-24 years old) and older workers (55-65 years old).

Indeed, while the concentration index of safe occupations among older workers is slightly above 1, safe

occupations are under-represented among the youngest workers (such results are consistent across all

13 It should be stressed here that age is of course not the only determinant of mortality risk from Covid-19, and that

other pre-existing medical conditions probably play an important role. Unfortunately, we do not observe individual

medical conditions in our data.

14 LFS data only report age as a categorical variable representing the midpoint of the 5-year age interval the individual

belongs to. This clearly affects the precision of our estimates for the average age.



27 countries; see Figure A B. 1 in Annex B). A similar pattern is detected if we consider specifically the

relative presence of young workers in occupations that are suitable for remote working. As young workers

are less likely to develop severe forms of the disease, their over-representation in epidemiologically

hazardous occupations could reduce the risk of job-related mortality during the pandemic.15

58. This concentration of young workers in unsafe jobs may be related to selection effects: many

occupations in category 1 (such as professionals in Business Administration) require high levels of skills,

and the more skilled individuals below age 24 are likely to be still in education. Another explanation is that

young workers at the very beginning of their career are involved in lower ranked, often front-office,

positions, involving frequent and risky contacts with customers. Indeed, the share of workers involved in

safe occupation is steadily increasing with age up to the age of 39, and then stabilises at about 54%

(Figure 11). Interestingly, this flattening in the age profile of the exposure to epidemiological risk is the

by-product of a decline at older ages of the share of jobs that can be carried out in remote, and an increase

of those that involve limited interactions, either with co-workers or customers.

15 Clearly being a young worker does not automatically makes the exposure to risky jobs worthwhile. However, an

allocation of jobs where younger individuals are employed in riskier occupations is likely to drastically reduce mortality.

Indeed, when riskier jobs are essential for the functioning of the economy and need to be performed, the best allocation

is one where individuals performing them are those with lower probability of dealing with more aggressive forms of

Covid-19. In this regard, the literature studying optimal lockdown policies often prescribes a differentiation in terms of

age (see Acemoglu et al. (2020[27]) and Ichino, Favero and Rustichini (2020[28]). Moreover, it is worth stressing that

here we are not considering spillover effects, notably the possibility of young workers transmitting the disease to more

vulnerable members of their family or social network. This is of course an important issue to consider in the design of

policies to contain the pandemic, but is something that we are not in a position to analyse with the available data.



Figure 11. Overall shares of jobs by category and age

Note: The figure shows the percentage of workers in the different job categories of our taxonomy, by age group. Pooled data from the 27

countries of the sample. Data refer to 2018.


59. In 19 countries out of 27 (Figure A B. 2 in Annex B) older workers are under-represented among the

jobs that can be done from home. This can be also explained by the lower level of proficiency of older adults

in the use of digital devices (OECD, 2015[20]).

60. Women are over-represented at the two ends of the work-safety ladder, as they are more likely to

be employed both in jobs that can be carried out remotely and in unsafe jobs (Figure 12). This indicates inter

alia that it could be fairly misleading to confine the definition of safe jobs to those that can be carried out from

home, as done by most of the literature reviewed in Section II. The over-representation of women in unsafe

jobs involves all countries, with the exception of Cyprus16, Greece and Romania (Figure A B. 3 in Annex B).17

16 A. Note by Turkey: The information in this document with reference to “Cyprus” relates to the southern part of the

Island. There is no single authority representing both Turkish and Greek Cypriot people on the Island. Turkey

recognises the Turkish Republic of Northern Cyprus (TRNC). Until a lasting and equitable solution is found within the


B. Note by all the European Union Member States of the OECD and the European Union: The Republic of Cyprus is

recognised by all members of the United Nations with the exception of Turkey. The information in this document relates

to the area under the effective control of the Government of the Republic of Cyprus.

17 This is partly due to the fact that occupations that are traditionally women-dominated, like nurses and primary school

and early childhood teachers, feature unsafe jobs only.



Figure 12. Concentration indexes of job categories by gender

Note: The figure shows concentration indexes for job categories by gender. Concentration indexes are computed as the ratio between the share

of jobs of category i for gender j over the share of category i in total employment, pooling data from the 27 countries of the sample. Numbers

greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Vertical bars

measure one standard deviation above and below the cross-country average of concentration indexes. Data refer to 2018.


61. There is instead a monotonic relationship between educational attainments of the workforce and

exposure to epidemiological risk (Figure 13). Low-educated workers are largely over-represented in unsafe

jobs.



Figure 13. Concentration indexes of job categories by education

Note: The figure shows concentration indexes for job categories by education level. We rely on LFS threefold categorisation derived from

ISCED2011 (low: lower secondary, middle: upper secondary, high: higher education attainment). Concentration indexes are computed as the

ratio between the share of jobs of category i for education level j over the share of category i in total employment, pooling data from the 27

countries of the sample. Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that

specific category. Vertical bars measure one standard deviation above and below the cross-country average of concentration indexes. Data

refer to 2018.


Hazardous jobs and economic vulnerability

62. A better assessment of the economic vulnerability of the workers most likely to be made redundant

in case of a long lasting epidemiological risk may come by analysing endogenous (to the labour market)

characteristics such as incomes, job security, and under-employment.

63. Figure 14 reports concentration indices by quintile of the earning distribution. The focus here is on

dependent employment. Workers holding safe jobs are seriously under-represented at the bottom of the

distribution. This is especially true for jobs in category 1, whose prevalence markedly increases as we

move toward the upper quintiles of the earnings distribution.



Figure 14. Concentration indexes of job categories by income quintiles

Note: The figure shows concentration indexes for job categories by income quintile. Concentration indexes are computed as the ratio between

the share of jobs of category i for quintile of income j over the share of category i in total employment, pooling data from 20 countries of the

sample. Data on income for Austria, the Czech Republic, Finland, Iceland, Norway, Spain and Sweden are not available. Numbers greater

(lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Vertical bars measure one

standard deviation above and below the cross-country average of mean of the country-specific concentration indexes. Data refer to 2018.


64. We find a strong concentration of unsafe jobs in families of more than five members (Figure 15).

Incidentally, large families may be at a serious disadvantage also when carrying out safe jobs: home is

arguably a very poor substitute to the office if a large family lives in a small apartment, especially if children

or other family members require assistance by those staying at home (e.g. because of school closures).



Figure 15. Concentration indexes of job categories by household size

Note: The figure shows concentration indexes for job categories by household size. Concentration indexes are computed as the ratio between

the share of jobs of category i for household size j over the share of category i in total employment, pooling data from the 27 countries of the

sample. Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category.

Vertical bars measure one standard deviation above and below the cross-country average of concentration indexes. Data refer to 2018.


65. Finally, epidemiological risk appears to be positively correlated with unemployment risk. Job losses

are typically concentrated on fixed-term contracts, especially in European countries with strong

employment protection. These contracts are over-represented in the pool of unsafe jobs (Figure 16), and

such evidence is consistent across examined countries.18 Part-timers are also disproportionally involved

(Figure 17) and they are frequently under-employed (27.2% of part-time was involuntary in the EU27 even

in the buoyant labour market conditions of 2018). In the United States solo self-employed are also over-

represented in the pool of unsafe jobs (Figure 18). Thus, it should not come as a surprise that unsafe jobs

are more prevalent among workers with relatively short tenures (i.e., having been for less than 6 months

in the current job, Figure 19).

18 Concentration indexes for unsafe jobs among workers with a temporary contract are above 1 for all 26 countries

analysed. Austria, Switzerland and Germany report the lowest figures (around 1.1), whereas Romania and Estonia

the highest (around 1.5).



Figure 16. Concentration indexes of job categories by type of contract

Note: The figure shows concentration indexes for job categories by contract type. Concentration indexes are computed as the ratio between the

share of jobs of category i for type j over the share of category i in total employment, pooling data from 26 countries of the sample (data are not

available for the US). Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that

specific category. Vertical bars measure one standard deviation above and below the cross-country average of concentration indexes. Data

refer to 2018.




Figure 17. Concentration indexes of job categories by working time arrangement

Note: The figure shows concentration indexes for job categories by working-time arrangement. Concentration indexes are computed as the ratio

between the share of jobs of category i for arrangement j over the share of category i in total employment, pooling data from the 27 countries of

the sample. Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific

category. Vertical bars measure one standard deviation above and below the cross-country average of concentration indexes. Data refer to

2018.




Figure 18. Concentration indexes of job categories by type of self-employment in the US

Note: The figure shows concentration indexes of job categories by type of self-employment. Concentration indexes are computed as the ratio

between the share of jobs of category i for type of self-employment j over the share of category i in total employment. Data refer to US in 2018

(data are not available for EU countries). Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-

representation) in that specific category.

Source: Current Population Survey (CPS).



Figure 19. Concentration indexes of job categories by tenure

Note: The figure reports concentration indexes of job categories for workers with short tenure (entered the current employment over the last 6

months). Concentration indexes are computed as the ratio between the share of jobs of category i among workers with short tenure over the

share of category i in total employment, pooling data from 26 countries of the sample (data are not available for the US). Numbers greater

(lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Vertical bars measure one

standard deviation above and below the cross-country average of concentration indexes. Data refer to 2018.


Targeting and reallocation

66. Overall, there is no evidence that COVID-19 acts as a Great Leveller like the Black Death, the

Russian revolution, and the World Wars (Scheidel, 2018[21])). Job-related epidemiological risk has a very

uneven distribution across sectors, occupations and firms. It also involves a rather specific worker profile,

broadly corresponding to the same characteristics that even in normal times are associated with a high risk

of job loss. These workers are then at a double disadvantage, as they might have lower chances of finding a

new job following a pandemic-induced layoff twice vulnerable because their job losses can be persistent as

they had already a weak labour market position before the crisis.

67. The fact that SMEs are hit particularly hard in this recession, compared for instance with the Great

Recession, poses a major challenge to labour market policies aimed at stabilizing employment over the cycle

as these policies are often not tailored to the small business sector and the solo self-employed. Furthermore,

plant closures, more frequent among small firms, would destroy at the same time firm-specific human,

physical and relational capital.

68. There is a core group of workers who are particularly exposed to long spells of joblessness and

labour market related hardship. This group is mainly composed of individuals with low levels of education,

currently employed in small units and performing unsafe occupations in the two sectors mentioned at the



outset (arts, entertainment and recreation, and accommodation and food service activities). We estimate that

in the countries covered by our analysis there are 2.0 million such individuals, representing 0.5% of the total

workforce and 6.7% of the total employment in the two aforementioned sectors.19

69. These very vulnerable groups, as the other holders of unsafe jobs, are distributed rather uniformly

over the age distribution above the prime age, and involve typically a higher share of very young people

(below the age of 25).20 Hence, early retirement is not an option, even though it can appear tempting from a

political point of view in countries with a large share of older voters.

70. In a number of countries, the policy response has been to extend the standard policy tools used to

contain job losses – employment protection, and short-time work – to small firms. A few countries (e.g.,

Greece, Spain and Italy), have banned economic layoffs in all kind of firms. In almost all European countries,

the coverage of STW has been extended to small business while funding subsidised working time reductions

via general government revenues.21 These measures were necessary during the lockdown, but can only be

temporary. If protracted over time, they would hinder worker reallocation.

71. Reallocation is even more important when account is made of the increasing number of people put

at the margin of the labour market by the hiring freeze and the collapse of new business start-ups. STW is

more costly than unemployment benefits because not only it typically offers higher replacement rates than

unemployment benefits, but it also involves moral hazard problems that can be particularly serious in the

case where there is no experience-rating and small firms are involved that operate in sectors not directly

exposed to the pandemic risk.

72. Serious consideration should therefore be given to: i) better targeting the policies to the sectors,

occupations and firms most hardly hit by the crisis; and ii) devising policies, such as combinations of STW

and wage insurance22, that could encourage the mobility of the workers that are twice vulnerable under the

pandemic towards those occupations and sectors that may offer greater employment opportunities.

73. Also, given that the activities that workers reallocated to different firms and sectors will be performing

are likely to differ from those in their previous occupations, offering training courses could help make the

transition smoother. In a context of support for this reallocation, policies could also encompass hiring

incentives for firms able to absorb workers released by unsafe or less productive sectors.

74. Most of the job creators – notably those related to the health value chain – could involve mainly

skilled workers, hence may not offer employment opportunities to the twice vulnerable groups described

above. There may be employment opportunities even for unskilled workers in essential activities and in new

disinfection-related jobs created with the goal of containing the pandemic. The problem is that these jobs

may carry with them significant health risk and offer relatively low wages, and hence may not be particularly

appealing even to the long-term unemployed as discussed in the next section.

19 Since information on firm size are not available for the US, we assume that the EU share of workers in firms with

less than 20 employees among low-educated workers in unsafe jobs in the two aforementioned sectors is equal to the

US one for the same set of individuals. 20 See Figure A B. 5 in Annex B.

21 See OECD (2020[18]) for a review of the economic initiatives undertaken so far by different countries.

22 With “wage insurance” we refer to a measure complementing the wage of workers accepting to move to sectors

offering lower wages compared to those of the initial employment.



Essential and non-essential jobs: How risky? Who does them?

75. In order to identify those jobs that can be considered essential, we use the taxonomy provided in

Fasani and Mazza (2020[14]), which identifies the occupations that need to be performed even during a

pandemic in order to keep citizens healthy, safe and fed. The list of such “key” occupations can be found in

Table A A. 19 in Annex A.

76. Essential occupations employ 119.5 million workers in our sample, representing 32.5% of the total

employment.23 About 60% of these essential workers (roughly 70 million people) hold a job that we classify

as unsafe. Indeed, as shown by Figure 20, unsafe jobs are over-represented in essential occupations.

Norway is the country with the highest share of unsafe jobs among essential workers (66%), whereas

Romania is the lowest (35%). Vice versa, safe jobs, notably activities that can be carried out remotely, are

severely under-represented in essential occupations. We estimate that only 19% of essential workers have

a job that can be performed remotely.

Figure 20. Concentration indexes of job categories in essential occupations

Note: The figure reports concentration indexes of job categories in essential occupations. Concentration indexes are computed as the ratio

between the share of jobs of category i in occupations considered essential over the share of category i in total employment, pooling data from

the 27 countries of the sample. Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in

that specific category. Vertical bars measure one standard deviation above and below the cross-country average of concentration indexes. Data

refer to 2018.


23 Information at the country level can be found in Table A A. 20 in Annex A.



77. These jobs often offer low wages, and employ disproportionally migrant workers who have typically

lower reservation wages than natives.24 Migrants are indeed over-represented among both essential

occupations and unsafe jobs (Figure 21). As shown by Fasani and Mazza (2020[14]), occupations like

cleaners and helpers, mining and construction, machine and food processing operators are often

dominated by migrants.

Figure 21. Concentration indexes of job categories by nativity status

Note: The figure shows concentration indexes for job categories by country of origin. Concentration indexes are computed as the ratio between

the share of jobs of category i for status j over the share of category i in total employment, pooling data from the 27 countries of the sample.

Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Vertical

bars measure one standard deviation above and below the cross-country average of concentration indexes. Data refer to 2018.


78. To attract more workers (and in particular young workers who are less exposed to the risk of severe

forms of COVID-19), under conditions in which migration is restricted by tighter border controls, wages should

compensate for the higher epidemiological risk involved by these jobs, a risk that was not perceived before the

COVID-19 pandemic, as documented in the following section.

24 In the US, the only country for which we have data on ethnicity, black and other minority ethnic groups tend to be

over-represented in unsafe occupations. On the contrary, Asians tend to be over-represented in category 1 jobs while

white individuals perform mostly perform jobs in safe occupations, but in no specific category (see Figure A B. 8 in

Annex B).



79. The notion of “essential” occupations also allows for the identification of an additional group of highly

vulnerable workers, namely people holding unsafe jobs in non-essential occupations. These workers are in fact

likely to face a particularly high risk of layoff, as their jobs are presumably among the first to be affected by

lockdown measures, and among the last to be authorised to resume. According to our estimates, roughly 140

million workers in EU countries have a job in non-essential occupations, about two thirds of the total EU

employment; 58.5 million of them hold an unsafe job (27% of the total EU employment).

80. Similar to what we found for vulnerable workers considered in the previous section (i.e. individuals with

low education levels, in small firms in arts, entertainment and recreation, construction and accommodation and

food service activities), the traits associated with labour market hardship are over-represented among workers

with unsafe jobs in non-essential activities. Figure 22 shows that being an immigrant, having a temporary

contract and having a low level of education are predominant characteristics among these workers.25 Not

surprisingly, accommodation and food service activities, commercial and wholesale trade, construction, other

services, and arts, entertainment and recreation sectors are over-represented among holders of unsafe jobs in

non-essential occupations.26

Figure 22. Concentration indexes of workers’ characteristics in unsafe non-essential occupations

Note: The figure shows concentration indexes of individual characteristics for workers holding unsafe jobs in non-essential occupations (as defined

by Fasani and Mazza, 2020). Concentration indexes are computed as the ratio between the share of workers with a specific trait among holders of

unsafe jobs in non-essential occupations over the share of workers with that specific trait in total employment, pooling data from the 26 countries of

the sample (data not available for the United States). Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-

representation) for that specific characteristic. Vertical bars measure one standard deviation above and below the cross-country average of

concentration indexes. Data refer to 2018.


25 The age distribution of these workers, however, does not show relevant differences compared to the economy-wide

distribution. See Figure B.6 in Annex B.

26 See Figure B.7 in Annex B.



A risk-premium in unsafe jobs?

81. In a perfect labour market – where both workers and firms are price takers – in presence of a double

heterogeneity (workers having different views as to their preferred income-risk combination and firms using

different technologies) equilibrium wages compensate for differences in job-related health risk. More risk averse

workers would accept lower wages only in exchange to a lower exposure to job related health risk. The opposite

would happen for less risk-averse workers, as they would accept risky jobs insofar as they are paid more than

the other jobs.

82. There is therefore a matching of workers and firms in either relatively low pay and low risk jobs or in

high pay and high risk positions. These equilibrium compensating wage differentials involve wage premia

allotted as a reward for workers facing higher job related epidemiological risk.

83. The above holds insofar as both workers and firms are perfectly informed about risk. Absent the

perception of risk among workers, they would all rank best paid jobs above all other jobs. In other words, there

will not be the double heterogeneity (on the supply and demand side) required to have an equilibrium with

compensating wage differentials. The equilibrium distribution of wages will be degenerate even if firms use

different technologies.

84. We may think of the situation before the outbreak of the COVID-19 pandemic as one where the

perception of epidemiological risk was absent among workers and employers. Under these conditions, we

should not expect to observe a premium on unsafe jobs. Actually, some features that increase the

epidemiological risk of jobs – such as frequent personal contacts with a heterogeneous and dynamic crowd –

may actually be valued by some workers and preferred over jobs that are largely carried out in isolation.

85. The tables below reproduce estimates of the risk premium (the coefficient associated to jobs involving

epidemiological risk according to our definition) in 2018 (i.e. well before COVID-19, when the epidemiological

risk was likely not perceived by workers). We measure the riskiness of the job with the same index we use to

identify jobs belonging to category 3, i.e. our broadest definition of “safe” jobs. As LFS data allow for a limited

breakdown of occupations and use ISCO codes, our index is not a binary variable, but it takes values between

0 and 1 depending on the share of safe jobs in any given occupation belonging to the 3-digit ISCO classification

(in the EU LFS data) and to the more granular SOC classification of US data (for CPS regressions). Higher

values of the index denote less exposure to risk in that occupation.

86. As the harmonised EU LFS do not contain information on wages, we first run our regression on CPS

data for the United States (Table 1), and then we replicate the same exercise using data from the OECD

Programme for the International Assessment of Adult Competencies (PIAAC), covering 21 countries that

participated in the survey in 2011-12 (Table 2). As in the EU LFS data, PIAAC classifies occupations using

ISCO.

87. We find no wage premium for risky jobs, but rather a premium on safe jobs (or a discount on unsafe

jobs). The premium narrows down as we add more covariates, i.e., as we make jobs more comparable along

the epidemiological risk dimension. This can be seen by comparing in Table 1 the coefficient for safe job in the

baseline specification (first column), to the specification with Mincer-type controls (education and tenure, second

column) and to the specification with an extended number of controls (third column).

88. The additional covariates are generally statistically significant and with a sign in line with a priori

expectations: for instance, being a woman, being an immigrant, having a part-time job are all traits connected

to lower earnings. This suggests that we are improving our specification and that there may be other

characteristics (not captured by our data) that could explain why unsafe jobs appear to be paid less than safe

jobs even when we use extended controls. It should also be stressed that results for the US are consistent

between the two datasets employed (CPS and PIAAC), with the coefficients on the main variable of interest

that are always positive and statistically significant.



89. Asian countries have been more exposed to epidemiological risk before COVID-19. Hence, in these

countries there can be more awareness of the risks associated to having jobs involving frequent personal

contacts with a large and varying number of people. As PIAAC data cover also Japan and South Korea, we can

see that even in Asian countries we observe a wage premium on safe jobs, which is robust across the different

specifications.

90. A possible interpretation of these results is that, in addition to lack of information about epidemiological

risk, unsafe jobs are more prevalent among workers with a relatively low bargaining position (and employers

with monopsony power). This is consistent with our finding about the over-representation of migrants, temporary

workers, solo self-employed, and low educated individuals in unsafe jobs. Moreover, the match between safe

jobs and productivity-enhancing technologies could also contribute to explain these results.

Table 1. Wage premia on safe jobs – United States 2018

Baseline Controls Extended

Variables Log(weekly earnings) Log(weekly earnings) Log(weekly earnings) (1) (2) (3)

Safe job 0.514*** 0.324*** 0.148***

(0.0903) (0.0584) (0.0360)

Age 0.0828*** 0.0339***

(0.00555) (0.00450)

Age sq -0.000838*** -0.000308***

(5.78e-05) (4.92e-05)

Education (middle level) -0.172 -0.0109

(0.118) (0.0975)

Education (high level) 0.647*** 0.0976

(0.151) (0.117)

Foreign-born -0.0832***

(0.0118)

Area 1 0.0529***

(0.00710)

Area 2 0.139***

(0.00957)

Female -0.170***

(0.0135)

Part-time -0.927***

(0.0244)

Constant 6.329*** 4.142*** 5.337***

(0.0671) (0.135) (0.132)

Observations 157,286 157,286 157,286

R-squared 0.078 0.273 0.481

Age # education NO YES YES

Age sq # education NO YES YES

No. children dummies NO NO YES

No. children # gender NO NO YES

Sector dummies NO NO YES

Note: Robust standard errors in parentheses, clustered at occupation level

*** p<0.01, ** p<0.05, * p<0.1

Area 1: between 100k and 1 mln residents; area 2: more than 1 mln residents

Source: Current Population Survey (CPS, 2018).



Table 2. Wage premia on safe jobs – PIAAC data

Observations for specifications (1) and

(2)

Observations for specification (3)

Baseline Controls Extended

controls

Countries Safe job coeff

Safe job coeff

Safe job coeff

(1) (2) (3) (4) (5)

Austria 0.371*** 0.247*** 0.192*** 2,823 1,721

(0.0707) (0.0518) (0.0425)

Flanders (Belgium)

0.222*** 0.133*** 0.102*** 2,595 1,811

(0.0578) (0.0327) (0.0299)

Canada 0.381*** 0.213*** 0.106** 15,915 10,118

(0.100) (0.0699) (0.0503)

Czech Republic 0.273*** 0.187*** 0.128** 2,527 1,624

(0.0771) (0.0566) (0.0563)

Germany 0.458*** 0.271*** 0.137** 3,065 1,878

(0.0934) (0.0605) (0.0550)

Denmark 0.308*** 0.175*** 0.123*** 4,316 3,155

(0.0581) (0.0316) (0.0242)

Spain 0.256*** 0.118* 0.0230 2,338 1,464

(0.0936) (0.0614) (0.0504)

Estonia 0.397*** 0.320*** 0.231*** 3,877 2,895

(0.0905) (0.0824) (0.0569)

Finland 0.319*** 0.230*** 0.140*** 3,083 2,102

(0.0637) (0.0512) (0.0417)

France 0.324*** 0.189*** 0.153*** 3,544 2,516

(0.0748) (0.0561) (0.0453)

England/N. Ireland (UK)

0.521*** 0.406*** 0.301*** 4,639 2,880

(0.0879) (0.0635) (0.0540)

Ireland 0.319*** 0.179** 0.148** 2,623 1,536

(0.114) (0.0760) (0.0582)

Italy 0.286*** 0.147** 0.0782 1,673 901

(0.0877) (0.0632) (0.0582)

Japan 0.408*** 0.293*** 0.152*** 3,146 1,909

(0.110) (0.0901) (0.0509)

Korea 0.458*** 0.277*** 0.184** 2,999 1,905

(0.0965) (0.0753) (0.0804)

Netherlands 0.444*** 0.204*** 0.157*** 3,071 1,894

(0.0869) (0.0414) (0.0295)

Norway 0.309*** 0.201*** 0.100*** 3,075 2,104

(0.0668) (0.0414) (0.0303)

Poland 0.337*** 0.157** 0.0967 3,750 1,475

(0.121) (0.0651) (0.0700)

Slovak Rep. 0.346*** 0.237*** 0.143** 2,389 1,655

(0.0797) (0.0524) (0.0549)

Sweden 0.258*** 0.219*** 0.152*** 2,765 1,891

(0.0418) (0.0387) (0.0305)

United States 0.550*** 0.327*** 0.182*** 2,712 1,736 (0.111) (0.0782) (0.0595)



Note: Standard errors in parenthesis, clustered at the occupation level*** p<0.01, ** p<0.05, * p<0.1 The table reports three different

specifications of a Mincer-type wage regressions run using PIAAC data. The dependent variable is log hourly wage. Only the coefficient on the

main variable of interest "safe job", proxying for the share of safe jobs in each occupation, is reported. The specification reported in column (1)

does not include any additional regressors other than the constant. In column (2) we add age, age squared, education dummies, age-education

interactions, age squared-education interactions as controls. In column (3) we further add dummies for economic sector, gender, immigrant

status, part-time job, number of children, and number of children-gender interactions.

Source: Survey of Adult Skills (2011/12).

91. It is not yet possible to estimate comparable wage equations with more recent data that should

presumably incorporate the current awareness of epidemiological risk. Hence, at this stage, we can only

speculate on future developments. Should we expect that – as the epidemiological risk is fully perceived

by workers and firms – wages in unsafe jobs will increase relative to wages in safe jobs?

92. There is no doubt that workers (and employers) are informed about this risk by now. However,

workers may be in a weaker bargaining position than before COVID-19, given the current state of the

economy. In other words, the counterfactual is not the wage before the pandemic, but the wage under a

comparable recession. Relative to this counterfactual, wages should increase to induce workers to supply

labour. To match workers requests, firms will be required to provide higher wages and/or mitigate the

health risks by adopting safety and distancing measures, although such actions could have a negative

impact on productivity.

93. There is therefore the risk of a decline of employment and production capacity in essential goods

and services through a reduction of the supply of labour. Such risk would be even stronger the more

relatively elastic is final demand. To prevent this potential market disruption, public support in terms of

wage subsidies (reducing the wedge between labour costs for the employer and take-home pay) or wage

insurance (allowing workers to cumulate STW subsidies with wages in essential occupations) could be

warranted when targeted to these sectors. Information campaigns about safety standards and other

measures to mitigate health risk could also improve awareness among workers and employers and make

wages more responsive to the actual risk faced by workers. Ultimately, the public sector could intervene

directly into the market of essential goods and services whose productive capacity is about to be lost.

Workers training and reallocation

94. On-the-job training and retraining of unemployed adults will play a major role in mitigating the

negative effects of the pandemic on employment and productivity. Not only workers will move away from

unsafe non-essential jobs to safe (and unsafe) essential jobs, but the organisation of all jobs is likely to

change profoundly (Bloom and Prettner, 2020[22]).

95. While it is too early to assess the extent of these changes, it is highly likely that training in digital

skills should be required to ease this reallocation. Digitalisation will be pervasive beyond category 1 safe

jobs, where remote working is already in place: it will be important also among jobs that are unsafe under

current technologies as there will be the need to have less physical proximity to avoid contagion risks.

Thus, unlike in previous recessions, we have quite a good understanding of the skills that are required to

reduce job-related epidemiological risk. In particular, we know that proficiency in the use of digital devices

is essential for remote working. Importantly, increasing reliance on remote working would have the further

advantage of reducing mobility related health risk, which goes well beyond COVID-19: in many countries,

most work-related injuries occur while commuting to work.

96. PIAAC provides a valuable source of information on the training needs of the population in working

age and on the workers in more needs in different countries. It included an assessment of Problem Solving

in Technology-Rich Environment (PSTRE), aimed at evaluating the ability of adults to solve problems and

perform a wide range of tasks using digital devices (PIAAC Expert Group in Problem Solving in

Technology-Rich Environments, 2009[23]).



97. Adults taking the assessment are placed in 4 different proficiency levels, depending on their score:

Below Level 1, Level 1, Level 2 and Level 3.27 The conceptual framework allows to describe to some details

what adults at different levels are actually able to do. For instance, adults scoring Below Level 1 are only

able to perform tasks based on well-defined problems involving the use of only one function within a generic

interface to meet one explicit criterion without any categorical, inferential reasoning or transforming of

information.

98. Figure 23 shows that in all countries the share of workers scoring Below Level 1 is higher in unsafe

than in safe jobs.28 On average across countries, 37 percent of workers in unsafe jobs score Below Level

1, compared to 22 percent of workers holding safe jobs. A similar picture emerges if we focus on adults

with very low ICT skills (those without previous computer experience, or who failed the ICT test, or who

refused to take the computer assessment): on average across countries, 13 percent of workers employed

in safe occupations have very low ICT skills, compared to 23 percent of workers in unsafe jobs.29 This

evidence confirms the prior that re-training policies will need to target workers in unsafe jobs.30

Figure 23. Percentage of workers scoring below level 1 in PSTRE in safe and unsafe jobs

Note: The figure shows the share of workers in safe and unsafe occupations that score below Level 1 in the assessment of Problem Solving in

Technology-Rich Environments (PSTRE). Adults who did not receive a score in PSTRE because they had no previous computer experience, or

because they failed the ICT core assessment, or because they opted-out of the computer-based version of the assessment, are classified as

being Below Level 1. Safe jobs are those whose category 3 index is above 0.6.

Source: Survey of Adult Skills (2011/12)

27 Not all participants received a score in PSTRE. Adults who reported no previous experience with a computer, or

who failed a very elementary assessment of ICT skills were not administered the PSTRE assessment, and took a test

of literacy and numeracy (the two other domains assessed in PIAAC) using paper and pencil. Participants were also

given the possibility to simply opt out of the computer-based assessment. We classify all these cases (about 18% of

the overall sample across participating countries) at Below Level 1.

28 We classify as safe jobs all the occupations for which the index for category 3 is above 0.6. France, Italy and Spain

are missing from the graph because they did not administer the PSTRE assessment.

29 The shares are almost identical if we restrict the attention to essential occupations.

30 Due to budget constraints and organisational feasibility it is hard to devise an active labour market policy offered to

the universe of less-skilled workers. As long as there will be reallocation away from unsafe non-essential occupations

towards new essential jobs, it will be preferable to target the workers involved in this reallocation.



99. On the basis of pre-pandemic information, about 50 percent of jobs were carried out in ways that

would expose workers to significant risks of infections, and would therefore be considered “unsafe” during

a pandemic. Some of these jobs provide essential goods and services, and cannot be discontinued, even

at the peak of a pandemic wave. For this reason, all efforts should be made to make these jobs as safe as

possible.

100. The emergence of an epidemiological risk potentially poses an unprecedented problem of

reallocation. First, relative price adjustments and a decrease in demand for goods and services that pose

a higher epidemiological risk could cause a structural transformation, with a permanent shrinkage of certain

occupations and a growth in labour demand in other jobs or sectors (such as health, pharmaceutical and

digital technology industry). Additionally, under the current occupational structure, a new type of mismatch

has materialised. At-risk workers – such as the elderly and people with co-morbidities known to increase

the severity of the disease – would preferably hold jobs that can be carried out safely, while workers less

at risk – the youngsters and those without co-morbidities – could increasingly take up unsafe jobs,

especially in essential sectors. In order to encourage this latter group to take up essential but more unsafe

jobs, wages should offer a premium for epidemiological risk. As shown in this paper, this was not the case

before COVID-19, as this risk was not perceived and workers had a very low bargaining position. This will

be less likely in the future.

101. Labour supply will be impacted both because the risk of contagion is now well in the mind of

potential job seekers, but also because border restrictions enforced as a consequence of the pandemic

naturally limit mobility and labour supply. Wage subsidies targeted to firms offering essential services could

allow these compensating wage differentials to unfold without putting firms under serious strain.

102. About 60 percent of unsafe jobs are in non-essential occupations. Firms offering these jobs will

have to undergo major restructuring to reduce epidemiological risk. This may involve at least temporarily

sizeable productivity losses and a dramatic drop in labour demand. As our analysis suggests, most of the

workers involved in this restructuring had already a vulnerable position in the labour market before COVID-

19. Thus, policies should target twice-vulnerable workers who are at a high risk of labour market related

hardship.

103. Early retirement does not seem to be an option as these workers are spread all over the age

distribution. Yet, for those workers who are close to the retirement age, an extended duration of

unemployment benefits could provide a sort of bridging scheme to retirement at least until an effective

vaccine will be discovered and adopted.

104. The immediate policy response in most EU countries was to extend and facilitate the access to

short-time work schemes (STW). Such measures, which allow workers to keep their job (and the right to

be reinstalled) while suffering hours (and salary) cuts, are a good way to preserve productive matches in

the midst of an economic crisis (Boeri et al., 2011[24]; Cahuc, Kramarz and Nevoux, 2018[25]; Giupponi and

Landais, 2020[26]). However, they also hinder reallocation, as workers are usually not allowed to work while

receiving the benefits. Workers on STW should instead be allowed to take up temporarily jobs in essential

occupations – at least those in the private sector – without losing the option to go back to their original job

when the emergency is over (Giupponi and Landais, 2020[17]). More generally, STW should become as

5 Concluding Remarks



much as possible a wage insurance scheme, encouraging workers to take-up jobs paid less than their

previous job. This will also encourage young workers to take up jobs in essential activities.

105. Publicly provided general training could also target these twice-vulnerable workers. In the current

juncture, we have a better idea of the training needs than under previous recessions. There is also a better

understanding among workers of the benefits associated to gaining the option to carry out some activities

in remote. Nonetheless, it may be useful to establish that provision of income support in terms of STW or

unemployment benefits is conditional on attendance to training courses aimed at increasing digital

proficiency, which we have shown to be significantly lower for workers more vulnerable to the

consequences of the COVID-19 pandemic.



References

Acemoglu, D. et al. (2020), “Optimal Targeted Lockdowns in a Multi-Group SIR Model”, NBER

Working Paper, No. 27102, National Bureau of Economic Research, Cambridge, MA,

http://dx.doi.org/10.3386/w27102.

[27]

Adams-Prassl, A. et al. (2020), “Inequality in the Impact of the Coronavirus Shock: Evidence

from Real Time Surveys”, Discussion Paper Series, No. 13183, IZA, Bonn, http://www.iza.org

(accessed on 27 June 2020).

[8]

Barbieri, T., G. Basso and S. Scicchitano (2020), “Italian workers at risk during the Covid-19

epidemic”, Occasional Papers, No. 569, Bank of Italy.

[7]

Barrero, J., N. Bloom and S. Davis (2020), “COVID-19 Is Also a Reallocation Shock”, BFI

Working Paper, No. 2020-59, BFI, https://oui.doleta.gov/unemploy/claims_arch.asp.


[11]

Bloom, D. and K. Prettner (2020), COVID-19 and the macroeconomic effects of automation,

VoxEU.org, https://voxeu.org/article/covid-19-and-macroeconomic-effects-automation


[22]

Boeri, T. et al. (2011), “Short-time work benefits revisited: some lessons from the Great

Recession”, Economic Policy, Vol. 26/68, pp. 699-765.

[24]

Boeri, T., A. Caiumi and M. Paccagnella (2020), “Mitigating the work-safety trade-off”, Covid

Economics Vetted and Real-Time Papers 2, pp. 60-66, https://portal.cepr.org/call-papers-


[2]

Borjas, G. and H. Cassidy (2020), “The Adverse Effect of the COVID-19 Labor Market Shock on

Immigrant Employment”, NBER Working Paper, No. 27243, National Bureau of Economic

Research, Cambridge, MA, http://dx.doi.org/10.3386/w27243.

[12]

Cahuc, P., F. Kramarz and S. Nevoux (2018), “When Short-Time Work Works”, CEPR

Discussion Paper, No. 13041, Centre for Economic Policy Research, London,

https://cepr.org/active/publications/discussion_papers/dp.php?dpno=13041.

[25]

Chetty, R. et al. (2020), How Did COVID-19 and Stabilization Policies Affect Spending and

Employment? A New Real-Time Economic Tracker Based on Private Sector Data,

https://opportunityinsights.org/paper/tracker/ (accessed on 27 June 2020).

[16]

Dingel, J. and B. Neiman (2020), “How Many Jobs Can be Done at Home?”, BFI White Paper,

BFI, https://github.com/jdingel/DingelNeiman-workathome. (accessed on 27 June 2020).

[1]

Fasani, F. and J. Mazza (2020), “Immigrant Key Workers: Their Contribution to Europe’s COVID-

19 Response”, Policy Paper, No. 155, IZA, Bonn, http://www.iza.org (accessed on

24 June 2020).

[14]

Foucault, M. and V. Galasso (2020), “Working during COVID-19: Cross-Country Evidence from

Real-Time Survey Data”, OECD Social, Employment and Migration Working Papers, Vol. No.

246, https://www.oecd-ilibrary.org/social-issues-migration-health/oecd-social-employment-

and-migration-working-papers_1815199x.

[15]



Giupponi, G. and C. Landais (2020), Building effective short-time work schemes for the COVID-

19 crisis, VoxEU.org, https://voxeu.org/article/building-effective-short-time-work-schemes-

covid-19-crisis (accessed on 27 June 2020).

[17]

Giupponi, G. and C. Landais (2020), “Subsidizing Labor Hoarding in Recessions: The

Employment & Welfare Effects of Short Time Work”, CEPR Discussion Papers, No. 13310,

Centre for Economic Policy Research,

https://cepr.org/active/publications/discussion_papers/dp.php?dpno=13310 (accessed on

27 June 2020).

[26]

Gottlieb, C., J. Grobovšek and M. Poschke (2020), “Working from home across countries”, Covid

Economics Vetted and Real-Time Papers 8.

[6]

Hensvik, L., T. Le Barbanchon and R. Rathelot (2020), “Which Jobs Are Done from Home?

Evidence from the American Time Use Survey”, IZA Discussion Paper, No. 13138, IZA, Bonn,

https://github.com/tlebarbanchon/home- (accessed on 27 June 2020).

[3]

Ichino, A., C. Favero and A. Rustichini (2020), “Restarting the Economy While Saving Lives

Under Covid-19”, Discussion Paper, No. 14664, CEPR, London, UK, http://www.cepr.org

(accessed on 18 July 2020).

[28]

Lewandowski, P. (2020), “Occupational Exposure to Contagion and the Spread of COVID-19 in

Europe”, IZA Discussion Paper, No. 13227, IZA, Bonn, http://www.iza.org (accessed on

27 June 2020).

[9]

Mongey, S., L. Pilossoph and A. Weinberg (2020), “Which Workers Bear the Burden of Social

Distancing Policies?”, Working Paper, No. 2020-51, BFI, https://bfi.uchicago.edu/wp-

content/uploads/BFI (accessed on 27 June 2020).

[4]

OECD (2020), “Distributional risks associated with non-standard work: Stylised facts and policy

considerations”, Tackling Coronavirus Series, OECD Publishing, Paris,

http://www.oecd.org/coronavirus/policy-responses/distributional-risks-associated-with-non-

standard-work-stylised-facts-and-policy-considerations-68fa7d61/ (accessed on

18 July 2020).

[13]

OECD (2020), OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis,

OECD Publishing, http://dx.doi.org/10.1787/1686c758-en.

[18]

OECD (2015), Adults, Computers and Problem Solving: What’s the Problem?, OECD Skills

Studies, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264236844-en.

[20]

PIAAC Expert Group in Problem Solving in Technology-Rich Environments (2009), “PIAAC

Problem Solving in Technology-Rich Environments: A Conceptual Framework”, OECD

Education Working Papers, No. 36, OECD Publishing, Paris,

https://dx.doi.org/10.1787/220262483674.

[23]

Poletti, P. et al. (2020), Probability of symptoms and critical disease after SARS-CoV-2 infection,

http://arxiv.org/abs/2006.08471 (accessed on 27 June 2020).

[19]

Saltiel, F. (2020), “Who can work from home in developing countries?”, Covid Economics Vetted

and Real-Time Papers 6.

[5]



Scheidel, W. (2018), The Great Leveler: Violence and the History of Inequality from the Stone

Age to the Twenty-First Century, Princeton University Press,

https://press.princeton.edu/books/paperback/9780691183251/the-great-leveler (accessed on

27 June 2020).

[21]

Yasenov, V. (2020), “Who Can Work from Home?”, IZA Discussion Paper, No. 13197, IZA,

Bonn, http://www.iza.org (accessed on 27 June 2020).

[10]



Annex A. Statistical Annex31

Table A A. 1. Coefficients of job categories by ISCO 3-digit code

ISCO code

ISCO name Cat. 1

Cat. 2

Cat. 3

“Unsafe jobs”

111 Legislators and Senior Officials 0.97 0.98 1.00 0.00

112 Managing Directors and Chief Executives 1.00 1.00 1.00 0.00

121 Business Services and Administration Managers 0.93 0.95 0.98 0.02

122 Sales, Marketing and Development Managers 0.88 1.00 1.00 0.00

131 Production Managers in Agriculture, Forestry and Fisheries 0.00 0.33 0.67 0.33

132 Manufacturing, Mining, Construction and Distribution Managers

0.29 0.64 1.00 0.00

133 Information and Communications Technology Services Managers

1.00 1.00 1.00 0.00

134 Professional Services Managers 0.75 0.76 0.96 0.04

141 Hotel and Restaurant Managers 0.00 0.00 0.13 0.87

142 Retail and Wholesale Trade Managers 1.00 1.00 1.00 0.00

143 Other Services Managers 0.76 0.87 0.98 0.02

211 Physical and Earth Science Professionals 0.50 1.00 1.00 0.00

212 Mathematicians, Actuaries and Statisticians 1.00 1.00 1.00 0.00

213 Life Science Professionals 0.71 0.82 0.94 0.06

214 Engineering Professionals (excluding Electrotechnology) 0.43 0.96 1.00 0.00

215 Electrotechnology Engineers 0.82 1.00 1.00 0.00

216 Architects, Planners, Surveyors and Designers 0.70 0.70 1.00 0.00

221 Medical Doctors 0.00 0.00 0.00 1.00

222 Nursing and Midwifery Professionals 0.00 0.00 0.00 1.00

223 Traditional and Complementary Medicine Professionals 0.00 0.00 0.00 1.00

224 Paramedical Practitioners 0.00 0.00 0.00 1.00

225 Veterinarians 0.00 0.00 0.00 1.00

226 Other Health Professionals 0.00 0.00 0.00 1.00

231 University and Higher Education Teachers 0.91 0.92 0.95 0.05

232 Vocational Education Teachers 0.06 0.06 0.06 0.94

233 Secondary Education Teachers 1.00 1.00 1.00 0.00

234 Primary School and Early Childhood Teachers 0.00 0.00 0.00 1.00

235 Other Teaching Professionals 0.48 0.48 0.48 0.52

241 Finance Professionals 1.00 1.00 1.00 0.00

242 Administration Professionals 0.76 0.76 0.93 0.07

243 Sales, Marketing and Public Relations Professionals 1.00 1.00 1.00 0.00

251 Software and Applications Developers and Analysts 1.00 1.00 1.00 0.00

252 Database and Network Professionals 1.00 1.00 1.00 0.00

261 Legal Professionals 1.00 1.00 1.00 0.00

262 Librarians, Archivists and Curators 0.87 0.87 1.00 0.00

263 Social and Religious Professionals 0.06 0.06 0.81 0.19

264 Authors, Journalists and Linguists 0.80 0.80 0.80 0.20

265 Creative and Performing Artists 0.57 0.57 0.57 0.43

311 Physical and Engineering Science Technicians 0.26 0.75 0.77 0.23

31 Data and other materials are available for download at http://www.frdb.org/page/data.

http://www.frdb.org/page/data



312 Mining, Manufacturing and Construction Supervisors 0.00 0.00 0.00 1.00

313 Process Control Technicians 0.00 0.12 0.29 0.71

314 Life Science Technicians and Related Associate Professionals 0.00 0.55 0.82 0.18

315 Ship and Aircraft Controllers and Technicians 0.00 0.10 0.10 0.90

321 Medical and Pharmaceutical Technicians 0.00 0.00 0.00 1.00

322 Nursing and Midwifery Associate Professionals 0.00 0.00 0.00 1.00

323 Traditional and Complementary Medicine Associate Professionals

0.00 0.00 0.00 1.00

324 Veterinary Technicians and Assistants 0.00 0.00 0.00 1.00

325 Other Health Associate Professionals 0.11 0.12 0.12 0.88

331 Financial and Mathematical Associate Professionals 0.95 0.97 1.00 0.00

332 Sales and Purchasing Agents and Brokers 1.00 1.00 1.00 0.00

333 Business Services Agents 0.46 0.57 0.87 0.13

334 Administrative and Specialized Secretaries 0.80 0.80 0.80 0.20

335 Government regulatory associate professionals 0.36 0.41 0.45 0.55

341 Legal, Social and Religious Associate Professionals 0.50 0.50 0.50 0.50

342 Sports and Fitness Workers 0.16 0.16 0.18 0.82

343 Artistic, Cultural and Culinary Associate Professionals 0.08 0.09 0.14 0.86

351 Information and Communications Technology Operations and User Support Technicians

1.00 1.00 1.00 0.00

352 Telecommunications and Broadcasting Technicians 0.08 0.41 0.86 0.14

411 General Office Clerks 1.00 1.00 1.00 0.00

412 Secretaries (general) 1.00 1.00 1.00 0.00

413 Keyboard Operators 1.00 1.00 1.00 0.00

421 Tellers, Money Collectors and Related Clerks 0.80 0.80 0.80 0.20

422 Client Information Workers 0.40 0.40 0.62 0.38

431 Numerical Clerks 1.00 1.00 1.00 0.00

432 Material recording and Transport Clerks 0.10 0.61 0.61 0.39

441 Other Clerical Support Workers 0.24 0.24 0.43 0.57

511 Travel Attendants, Conductors and Guides 0.05 0.05 0.05 0.95

512 Cooks 0.00 0.00 0.02 0.98

513 Waiters and Bartenders 0.00 0.00 0.00 1.00

514 Hairdressers, Beauticians and Related Workers 0.03 0.03 0.03 0.97

515 Building and Housekeeping Supervisors 0.00 0.77 1.00 0.00

516 Other Personal Services Workers 0.05 0.05 0.11 0.89

521 Street and Market Salespersons 0.20 0.20 0.20 0.80

522 Shop Salespersons 0.28 0.28 0.28 0.72

523 Cashiers and Ticket Clerks 0.00 0.00 0.00 1.00

524 Other Sales Workers 0.09 0.09 0.11 0.89

531 Child Care Workers and Teachers’ Aides 0.01 0.01 0.01 0.99

532 Personal Care Workers in Health Services 0.00 0.00 0.00 1.00

541 Protective Services Workers 0.00 0.00 0.09 0.91

611 Market Gardeners and Crop Growers 0.00 0.52 0.52 0.48

612 Animal Producers 0.00 0.81 0.81 0.19

613 Mixed Crop and Animal Producers 0.00 0.94 0.94 0.06

621 Forestry and Related Workers 0.00 0.69 0.80 0.20

622 Fishery Workers, Hunters and Trappers 0.00 0.77 0.77 0.23

631 Subsistence Crop Farmers 0.00 0.50 1.00 0.00

632 Subsistence Livestock Farmers 0.00 0.00 0.00 1.00

633 Subsistence Mixed Crop and Livestock Farmers 0.00 0.34 0.68 0.32

711 Building Frame and Related Trades Workers 0.00 0.00 0.14 0.86

712 Building Finishers and Related Trades Workers 0.00 0.16 0.30 0.70

713 Painters, Building Structure Cleaners and Related Trades Workers

0.00 0.13 0.13 0.87

721 Sheet and Structural Metal Workers, Moulders and Welders, and Related Workers

0.00 0.59 0.59 0.41

722 Blacksmiths, Toolmakers and Related Trades Workers 0.00 0.96 0.98 0.02



723 Machinery Mechanics and Repairers 0.00 0.62 0.65 0.35

731 Handicraft Workers 0.05 0.51 0.63 0.37

732 Printing Trades Workers 0.16 1.00 1.00 0.00

741 Electrical Equipment Installers and Repairers 0.00 0.03 0.04 0.96

742 Electronics and Telecommunications Installers and Repairers 0.00 0.01 0.41 0.59

751 Food Processing and Related Trades Workers 0.00 0.41 0.41 0.59

752 Wood Treaters, Cabinet-makers and Related Trades Workers 0.00 0.48 0.98 0.02

753 Garment and Related Trades Workers 0.00 0.56 0.63 0.37

754 Other Craft and Related Workers 0.00 0.00 0.14 0.86

811 Mining and Mineral Processing Plant Operators 0.00 0.62 0.62 0.38

812 Metal Processing and Finishing Plant Operators 0.00 0.66 0.66 0.34

813 Chemical and Photographic Products Plant and Machine Operators

0.00 0.84 0.99 0.01

814 Rubber, Plastic and Paper Products Machine Operators 0.00 0.70 0.70 0.30

815 Textile, Fur and Leather Products Machine Operators 0.00 0.43 0.43 0.57

816 Food and Related Products Machine Operators 0.00 0.73 0.73 0.27

817 Wood Processing and Papermaking Plant Operators 0.00 1.00 1.00 0.00

818 Other Stationary Plant and Machine Operators 0.00 0.95 0.96 0.04

821 Assemblers 0.00 0.19 0.19 0.81

831 Locomotive Engine Drivers and Related Workers 0.00 0.36 0.36 0.64

832 Car, Van and Motorcycle Drivers 0.00 0.00 0.00 1.00

833 Heavy Truck and Bus Drivers 0.00 0.00 0.73 0.27

834 Mobile Plant Operators 0.00 0.80 0.84 0.16

835 Ships’ Deck Crews and Related Workers 0.00 0.00 0.00 1.00

911 Domestic, Hotel and Office Cleaners and Helpers 0.00 0.40 0.40 0.60

912 Vehicle, Window, Laundry and Other Hand Cleaning Workers 0.00 0.97 0.97 0.03

921 Agricultural, Forestry and Fishery Labourers 0.00 0.09 0.19 0.81

931 Mining and Construction Labourers 0.00 0.01 0.01 0.99

932 Manufacturing Labourers 0.00 0.29 0.29 0.71

933 Transport and Storage Labourers 0.00 0.13 0.13 0.87

941 Food Preparation Assistants 0.00 0.00 0.00 1.00

961 Refuse Workers 0.00 0.00 0.00 1.00

962 Other Elementary Workers 0.01 0.07 0.08 0.92

Note: The table reports job category coefficients assigned to each ISCO 3-digit occupation. Each coefficient is a proxy of

the share of jobs that belong to the category under consideration within a given ISCO code. ISCO code 951 has been

dropped due to inconsistencies between ICP INAPP and O*NET data.

Source: O*NET database.

Table A A. 2. Overall shares of job categories by country

Country Cat. 1 In Cat. 2 but not in 1 In Cat. 3 but not in 2 “Unsafe”

Austria .324 .167 .054 .455

Belgium .337 .127 .065 .471

Croatia .269 .182 .068 .481

Cyprus .334 .115 .053 .498

Czech Rep .285 .207 .07 .438

Denmark .308 .128 .051 .513

Estonia .318 .164 .079 .439

Finland .318 .149 .064 .469

France .318 .15 .056 .476

Germany .316 .169 .059 .456

Greece .275 .155 .049 .521

Hungary .259 .191 .068 .482

Iceland .328 .113 .061 .498

Ireland .291 .127 .058 .524



Italy .304 .148 .057 .491

Latvia .29 .158 .071 .481

Luxembourg .435 .096 .076 .393

Netherlands .339 .101 .061 .499

Norway .318 .123 .06 .499

Portugal .275 .168 .074 .483

Romania .17 .315 .07 .445

Slovak Rep .222 .175 .07 .533

Spain .24 .136 .065 .559

Sweden .34 .122 .07 .468

Switzerland .359 .129 .064 .448

UK .355 .098 .071 .476

US .332 .107 .08 .481

Note: The table reports the share of workers holding a job in any of the job categories of our taxonomy for the 27 countries

of the sample. Data refer to 2018.


Table A A. 3. Concentration indexes of job categories for male workers by country

Country IC Cat. 1 IC Cat. 2 IC Cat. 3 IC “unsafe”

Austria .877 1.041 1.072 .914

Belgium .961 1.052 1.074 .917

Croatia .788 .987 1.033 .965

Cyprus .784 .891 .948 1.052

Czech Rep .825 1.024 1.064 .918

Denmark .964 1.11 1.125 .881

Estonia .843 1.004 1.068 .913

Finland .962 1.12 1.139 .842

France .881 1.013 1.046 .95

Germany .902 1.054 1.077 .908

Greece .836 .953 .994 1.006

Hungary .753 .996 1.048 .948

Iceland .933 1.075 1.09 .91

Ireland .876 1.055 1.078 .929

Italy .865 .996 1.031 .967

Latvia .797 .987 1.05 .946

Luxembourg .979 1.026 1.044 .931

Netherlands 1.029 1.118 1.118 .882

Norway 1.013 1.145 1.16 .84

Portugal .88 1.025 1.068 .928

Romania .788 .957 .993 1.009

Slovak Rep .779 1.025 1.077 .932

Spain .962 1.056 1.084 .934

Sweden .985 1.119 1.124 .859

Switzerland .958 1.039 1.053 .935

UK .989 1.086 1.099 .891

US .937 1.064 1.067 .927

Note: The table reports concentration indexes of job categories for male workers by country. For any of the 27 countries

of the sample, concentration indexes are computed as the ratio between the share of jobs of category i for male workers

over the share of category i in total employment. Numbers greater (lower) than one denote over-representation

(under-representation) in that specific category. Data refer to 2018.




Table A A. 4. Concentration indexes of job categories for female workers by country


Austria 1.142 .955 .919 1.097

Belgium 1.042 .94 .915 1.096

Croatia 1.249 1.018 .961 1.042

Cyprus 1.24 1.12 1.054 .946

Czech Rep 1.218 .97 .92 1.103

Denmark 1.039 .876 .858 1.135

Estonia 1.167 .994 .929 1.091

Finland 1.038 .869 .847 1.173

France 1.129 .985 .952 1.053

Germany 1.114 .94 .91 1.107

Greece 1.229 1.063 1.01 .99

Hungary 1.297 1.004 .942 1.062

Iceland 1.076 .912 .894 1.106

Ireland 1.144 .935 .912 1.08

Italy 1.184 1.004 .959 1.043

Latvia 1.2 1.016 .952 1.052

Luxembourg 1.023 .97 .949 1.079

Netherlands .968 .866 .866 1.134

Norway .984 .839 .818 1.182

Portugal 1.124 .975 .928 1.077

Romania 1.282 1.056 1.011 .987

Slovak Rep 1.279 .967 .904 1.084

Spain 1.042 .931 .9 1.079

Sweden 1.012 .87 .861 1.158

Switzerland 1.047 .953 .938 1.076

UK 1.014 .905 .889 1.122

US 1.071 .928 .924 1.082

Note: The table reports concentration indexes of job categories for female workers by country. For any of the 27 countries

of the sample, concentration indexes are computed as the ratio between the share of jobs of category i for female workers




Table A A. 5. Concentration indexes of job categories for young workers (15-24) by country


Austria .799 .868 .859 1.169

Belgium .635 .769 .764 1.265

Croatia .483 .741 .736 1.285

Cyprus .554 .63 .62 1.384

Czech Rep .667 .896 .879 1.155

Denmark .487 .544 .538 1.439

Estonia .774 .813 .781 1.28

Finland .475 .645 .652 1.394

France .714 .846 .832 1.185

Germany .763 .835 .818 1.217

Greece .607 .672 .666 1.307

Hungary .68 .851 .815 1.199

Iceland .442 .499 .496 1.508

Ireland .67 .672 .653 1.315

Italy .579 .748 .737 1.273

Latvia .814 .973 .919 1.087



Luxembourg .609 .672 .689 1.481

Netherlands .475 .568 .557 1.445

Norway .487 .608 .599 1.403

Portugal .705 .84 .799 1.215

Romania .659 .998 .944 1.07

Slovak Rep .694 .839 .824 1.154

Spain .671 .729 .707 1.231

Sweden .521 .613 .617 1.436

Switzerland .808 .844 .839 1.199

UK .682 .717 .702 1.328

US .548 .651 .671 1.356

Note: The table reports concentration indexes of job categories for young workers (15-24) by country. For any of the 27

countries of the sample, concentration indexes are computed as the ratio between the share of jobs of category i for young

workers over the share of category i in total employment. Numbers greater (lower) than one denote over-representation



Table A A. 6. Concentration indexes of job categories for older workers (55-65) by country


Austria .994 1.051 1.057 .932

Belgium 1.05 1.03 1.036 .96

Croatia .974 1.064 1.062 .933

Cyprus .955 1.007 1.028 .972

Czech Rep .916 .99 1.004 .995

Denmark .987 1.055 1.068 .936

Estonia .818 .909 .952 1.062

Finland .994 1.028 1.034 .962

France 1.016 1.032 1.031 .966

Germany .959 1.002 1.015 .982

Greece .884 1.088 1.084 .923

Hungary .811 .969 .983 1.019

Iceland 1 1.063 1.068 .932

Ireland .99 1.086 1.103 .906

Italy 1.112 1.06 1.061 .937

Latvia .817 .915 .944 1.06

Luxembourg 1.103 1.092 1.071 .891

Netherlands 1.021 1.05 1.056 .944

Norway 1.148 1.134 1.16 .84

Portugal .924 .971 .975 1.027

Romania .735 1.118 1.09 .888

Slovak Rep .919 1 1.015 .987

Spain .925 .995 1.011 .991

Sweden .974 1.009 1.019 .979

Switzerland .958 1.01 1.013 .984

UK .949 .998 1.021 .977

US 1.084 1.087 1.091 .902

Note: The table reports concentration indexes of job categories for older workers (55-65) by country. For any of the 27

countries of the sample, concentration indexes are computed as the ratio between the share of jobs of category i for older

workers over the share of category i in total employment. Numbers greater (lower) than one denote over-representation





Table A A. 7. Concentration indexes by quintile of income (sample weighted average)

Quintile IC Cat. 1 IC Cat. 2 IC Cat. 3 IC “unsafe”

Quintile 1 .488 .576 .603 1.428

Quintile 2 .642 .804 .823 1.191

Quintile 3 .964 1.016 1.017 .981

Quintile 4 1.229 1.164 1.146 .842

Quintile 5 1.666 1.444 1.356 .615

Note: The table reports concentration indexes of job categories by quintile of income, pooling data from 20 countries of the

sample. Data on income for Austria, the Czech Republic, Finland, Iceland, Norway, Spain and Sweden are not available. For

each country concentration indexes are computed as the ratio between the share of jobs of category i in quintile of income j




Table A A. 8. Concentration indexes for category 3 jobs by quintile of income and country

Country Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5

Belgium .679 .809 .964 1.127 1.42

Croatia .732 .861 .946 .96 1.212

Cyprus .741 .871 1.102 1.124 1.448

Denmark .63 .747 .982 1.133 1.515

Estonia .724 .765 .939 1.144 1.257

France .74 .834 .956 1.088 1.464

Germany .75 .835 .976 1.11 1.32

Greece .758 .858 1.008 1.027 1.228

Hungary .801 .873 .965 1.06 1.344

Ireland .624 .725 .901 1.107 1.361

Italy .684 .835 .996 1.059 1.279

Latvia .769 .8 .975 1.119 1.362

Luxembourg .654 .916 1.132 1.176 1.275

Netherlands .583 .798 .944 1.136 1.503

Portugal .671 .716 1 1.164 1.327

Romania .78 .814 .877 .926 1.041

Slovak Rep .557 .792 .949 1.131 1.368

Switzerland .822 .855 .929 1.04 1.293

UK .586 .721 .943 1.122 1.408

US .603 .823 1.017 1.146 1.356

Note: The table reports concentration indexes of category 3 jobs by quintile of income for 20 countries of the sample. Data

on income for Austria, the Czech Republic, Finland, Iceland, Norway, Spain and Sweden are not available. For any of the

20 countries for which data are available, concentration indexes are computed as the ratio between the share of jobs of

category 3 in quintile of income j over the share of category 3 in total employment. Numbers greater (lower) than one denote

over-representation (under-representation) in that specific category. Data refer to 2018.


Table A A. 9. Concentration indexes by job category and economic sector (sample weighted average)

Economic sector IC Cat.1 IC Cat.2 IC Cat.3 IC "unsafe"

Agriculture/Forestry/Fishing 0.144 1.252 1.266 0.698

Mining/Quarrying 0.831 1.368 1.312 0.665

Manufacturing 0.771 1.344 1.254 0.727

Utilities 1.219 1.215 1.232 0.752

Waste Management 0.670 0.797 0.941 1.065



Construction 0.458 0.617 0.752 1.268

Trade 1.131 1.011 0.931 1.077

Transportation/Storage 0.550 0.650 0.928 1.079

Accommodation and food 0.238 0.254 0.314 1.728

Information and communication 2.355 1.740 1.608 0.350

Financial and insurance act. 2.649 1.946 1.759 0.181

Real estate 1.029 0.971 1.426 0.543

Professional/Scientific/Techn. 2.239 1.792 1.655 0.296

Administrative and support act. 0.768 1.023 1.074 0.920

Public administration 1.360 1.080 1.070 0.922

Education 1.345 1.039 0.979 1.024

Health and social work act. 0.479 0.415 0.482 1.562

Arts/Entertainment 1.058 0.888 0.904 1.105

Other services 0.716 0.870 0.968 1.034

Households as employers 0.033 0.540 0.499 1.532

Extraterritorial bodies 1.851 1.416 1.360 0.582

Note: The table reports concentration indexes of job categories by economic sector (Nace rev 2), pooling data from the 27

countries of the sample. For any country concentration indexes are computed as the ratio between the share of jobs of

category i in economic sector j over the share of category i in total employment. Numbers greater (lower) than one denote

over-representation (under-representation) in that specific category. Data refer to 2018.


Table A A. 10. Concentration indexes of category 3 jobs by economic sector and country

Economic sector Austria Belgium Cyprus Czech Rep

Denmark Estonia Germany Spain Switzerland

Accommodation and food 0.385 0.291 0.406 0.523 0.253 0.33 0.386 0.247 0.359

Administrative and support act.

1.018 1.055 0.974 0.964 1.146 0.875 1.07 1.023 1.016

Agriculture/Forestry/Fishing 1.343 1.185 0.98 1.279 1.374 1.221 1.279 1.029 1.361

Arts/Entertainment 0.936 1.119 1.143 1.048 1.023 1.057 0.912 1 0.933

Construction 0.826 0.715 0.618 0.74 0.731 0.745 0.84 0.701 0.636

Education 1.024 0.962 1.068 0.89 0.869 0.863 1.072 1.118 1.049

Extraterritorial bodies 1.552 1.624 0.94 1.509 1.721 0.729 1.368 2.268 1.139

Financial and insurance act. 1.701 1.792 1.859 1.714 1.893 1.686 1.601 1.914 1.594

Health and social work act. 0.431 0.473 0.482 0.331 0.359 0.333 0.449 0.363 0.496

Households as employers 0.892 0.42 n/a 0.071 0.437 n/a 0.781 0.85 0.891

Information and communication

1.602 1.707 1.701 1.653 1.752 1.631 1.518 1.937 1.357

Manufacturing 1.233 1.244 1.241 1.094 1.423 1.125 1.213 1.342 1.12

Mining/Quarrying 1.217 1.009 1.337 1.1 1.745 1.075 1.287 1.379 1.022

Other services 0.738 0.922 0.496 0.585 1.01 0.722 0.813 0.703 0.917

Professional/Scientific/Techn. 1.492 1.639 1.831 1.536 1.702 1.578 1.546 1.798 1.415

Public administration 1.026 1.059 0.753 1.053 1.429 1.034 1.129 0.964 1.21

Real estate 1.541 1.482 1.817 1.528 1.632 1.185 1.46 1.868 1.516

Trade 0.914 0.953 0.948 0.879 0.91 1.021 0.866 0.946 0.942

Transportation/Storage 1 1.062 1.137 1.078 1.035 1.08 0.987 1.009 1.091

Utilities 1.187 1.314 1.243 1.253 1.741 1.118 1.215 1.506 0.964

Waste Management 0.982 1.042 0.576 1.137 1 1.08 0.98 1.007 1.009

Note: The table reports concentration indexes of category 3 jobs by economic sector a set of 9 countries belonging to our sample. For any

country concentration indexes are computed as the ratio between the share of jobs of category 3 in economic sector j over the share of category

3 in total employment. Numbers greater (lower) than one denote over-representation (under-representation) in that specific category. Data refer

to 2018.




Table A A. 11. Concentration indexes of job category 3 jobs by economic sector and country

Economic sector Croatia Finland France Greece Hungary Iceland Ireland Italy Latvia Luxemb.

Accommodation and food 0.326 0.23 0.378 0.288 0.355 0.504 0.328 0.265 0.414 0.306

Administrative and support act.

0.834 1.188 1.017 0.998 0.998 1.155 1.065 0.912 0.861 0.885

Agriculture/Forestry/Fishing 1.364 1.365 1.296 1.205 1.237 1.283 1.609 0.931 1.083 1.409

Arts/Entertainment 1.127 1.006 0.994 1.015 1.147 0.948 1.143 1.02 1.102 0.965

Construction 0.765 0.761 0.767 0.76 0.654 0.759 0.733 0.66 0.823 0.727

Education 0.809 0.94 1.011 1.098 0.772 0.717 1.078 1.083 0.736 0.6

Extraterritorial bodies 1.252 1.527 1.223 1.196 1.649 1.906 1.492 1.446 1.892 1.442

Financial and insurance act. 1.723 1.763 1.725 1.866 1.774 1.886 1.721 1.857 1.811 1.529

Health and social work act. 0.287 0.335 0.515 0.418 0.508 0.398 0.42 0.424 0.395 0.535

Households as employers 0.539 0.149 0.622 0.689 0.595 n/a 0.271 0.454 0.006 0.628

Information and communication

1.728 1.738 1.645 1.737 1.68 1.645 1.153 1.79 1.636 1.425

Manufacturing 1.279 1.318 1.219 1.299 1.149 1.167 1.258 1.244 1.247 1.152

Mining/Quarrying 1.341 1.363 1.332 1.196 1.324 1.365 1.29 1.295 1.162 1.262

Other services 0.615 0.923 0.784 0.718 0.544 1.129 0.786 0.666 0.499 0.855

Professional/Scientific/Techn.

1.607 1.593 1.573 1.908 1.724 1.645 1.723 1.731 1.64 1.458

Public administration 1.125 1.2 0.99 0.998 0.929 1.305 1.071 1.037 1.227 1.015

Real estate 1.497 1.629 1.601 1.912 1.585 1.608 1.628 1.53 0.732 1.364

Trade 0.915 0.891 0.956 0.858 0.946 0.926 0.979 0.98 1.019 0.921

Transportation/Storage 0.998 0.908 1.113 0.921 1.189 0.922 0.863 1.1 1.094 0.909

Utilities 0.925 1.512 1.273 1.322 1.347 1.341 1.298 1.291 1.056 1.186

Waste Management 0.834 1.399 1.048 0.764 1.183 1.026 1.139 0.78 0.902 1.105




to 2018.


Table A A. 12. Concentration indexes of job category 3 jobs by economic sector and country

Economic sector Netherlands Norway Portugal Romania Slovak Rep Sweden UK US

Accommodation and food 0.325 0.413 0.333 0.299 0.325 0.303 0.357 0.301

Administrative and support act. 1.002 1.152 0.926 0.53 0.835 1.086 1.111 1.148

Agriculture/Forestry/Fishing 1.259 1.473 1.474 1.456 1.411 1.419 1.191 1.256

Arts/Entertainment 1.054 0.95 1.017 0.968 1.251 1.024 1.046 0.782

Construction 0.792 0.834 0.687 0.569 0.782 0.782 0.885 0.728

Education 1.09 0.818 1.015 1.045 0.715 0.626 0.874 0.979

Extraterritorial bodies 1.946 1.491 0.915 n/a 1.362 1.695 1.036 n/a

Financial and insurance act. 1.739 1.92 1.807 1.623 1.949 1.823 1.662 1.794

Health and social work act. 0.575 0.277 0.431 0.249 0.272 0.321 0.521 0.516

Households as employers 0.589 n/a 0.743 0.6 0.173 0.011 0.845 0.048

Information and communication 1.727 1.756 1.749 1.571 1.846 1.737 1.677 1.468

Manufacturing 1.184 1.397 1.207 1.058 1.182 1.31 1.305 1.312

Mining/Quarrying 1.533 1.387 1.246 1.119 1.321 1.297 1.321 1.331

Other services 0.82 0.968 0.708 0.721 0.711 0.929 0.813 1.121

Professional/Scientific/Techn. 1.611 1.749 1.727 1.497 1.803 1.611 1.582 1.692

Public administration 1.259 1.375 0.967 0.951 0.955 1.299 1.097 1.077

Real estate 1.489 1.675 1.609 1.519 1.814 1.541 1.452 1.362

Trade 0.858 1.096 0.921 0.816 0.955 1.058 0.901 0.948

Transportation/Storage 1 1.062 1.089 0.829 1.193 1.002 0.859 0.803

Utilities 1.447 1.445 1.267 0.895 1.113 1.305 1.273 1.212



Waste Management 1.152 1.026 1.064 0.645 1.043 1.086 0.979 0.882




to 2018.


Table A A. 13. Concentration indexes of job categories by living area (sample weighted average)

Area IC Cat. 1 IC Cat. 2 IC Cat. 3 IC “unsafe”

Metropolitan 1.1 1.027 1.021 .977

Non-metropolitan .845 .957 .965 1.037

Note: The table reports concentration indexes of job categories by living area, pooling data from the 27 countries of the sample.

For any country concentration indexes are computed as the ratio between the share of jobs of category i in area j over the

share of category i in total employment. Metropolitan areas are defined as areas with more than 100,000 individuals. Numbers

greater (lower) than one denote over-representation (under-representation) in that specific category. Data refer to 2018.


Table A A. 14. Concentration indexes of job categories for metropolitan areas by country


Austria 1.167 1 1.007 .991

Belgium 1.095 1.037 1.036 .96

Croatia 1.476 1.124 1.125 .865

Cyprus 1.147 1.085 1.072 .928

Czech Rep 1.389 1.104 1.084 .893

Denmark 1.24 1.067 1.064 .94

Estonia 1.248 1.089 1.062 .92

Finland 1.305 1.094 1.073 .917

France 1.226 1.083 1.076 .916

Germany 1.146 1.021 1.024 .971

Greece 1.262 1.026 1.033 .969

Hungary 1.479 1.131 1.124 .867

Iceland 1.171 1.057 1.046 .954

Ireland 1.237 1.045 1.036 .968

Italy 1.22 1.073 1.057 .941

Latvia 1.169 1.054 1.029 .969

Luxembourg 1.469 1.298 1.254 .608

Netherlands 1.094 1.034 1.034 .966

Norway 1.28 1.122 1.106 .894

Portugal 1.255 1.072 1.044 .952

Romania 1.729 .942 .987 1.016

Slovak Rep 1.536 1.166 1.139 .878

Spain 1.188 1.059 1.043 .966

Sweden 1.229 1.097 1.083 .906

Switzerland 1.153 1.043 1.049 .94

UK 1.028 1.004 1.002 .998

US 1.042 1.016 1.01 .99

Note: The table reports concentration indexes of job categories for metropolitan areas by country. For any of the 27 countries

of the sample, concentration indexes are computed as the ratio between the share of jobs of category i in metropolitan areas

over the share of category i in total employment. Metropolitan areas are defined as areas with more than 100,000 individuals.

Numbers greater (lower) than one denote over-representation (under-representation) in that specific category. Data refer to 2018.




Table A A. 15. Concentration indexes of job categories for non-metropolitan areas by country


Austria .932 1 .998 1.003

Belgium .964 .987 .987 1.015

Croatia .758 .936 .935 1.07

Cyprus .821 .896 .912 1.088

Czech Rep .825 .955 .963 1.047

Denmark .87 .964 .967 1.032

Estonia .783 .921 .946 1.069

Finland .782 .933 .947 1.06

France .825 .933 .941 1.065

Germany .919 .988 .987 1.016

Greece .83 .982 .98 1.019

Hungary .765 .935 .939 1.065

Iceland .682 .896 .914 1.087

Ireland .857 .972 .981 1.017

Italy .872 .957 .967 1.034

Latvia .862 .957 .976 1.025

Luxembourg .866 .915 .928 1.112

Netherlands .876 .957 .956 1.044

Norway .881 .948 .956 1.044

Portugal .796 .946 .964 1.039

Romania .601 1.031 1.007 .992

Slovak Rep .859 .955 .962 1.033

Spain .771 .93 .947 1.042

Sweden .864 .943 .951 1.056

Switzerland .938 .981 .979 1.026

UK .963 .998 .997 1.003

US .717 .884 .923 1.083

Note: The table reports concentration indexes of job categories for non-metropolitan areas by country. For any of the 27 countries

of the sample, concentration indexes are computed as the ratio between the share of jobs of category i in non-metropolitan areas

over the share of category i in total employment. Non-metropolitan areas are defined as areas with less than 100,000 individuals.

Numbers greater (lower) than one denote over-representation (under-representation) in that specific category. Data refer to 2018.


Table A A. 16. Concentration indexes of category 3 jobs by level of education and country

Country Low Middle High

Austria 0.77 0.91 1.22

Belgium 0.75 0.84 1.23

Croatia 0.91 0.88 1.27

Cyprus 0.68 0.83 1.25

Czech Rep 0.68 0.92 1.28

Denmark 0.71 0.99 1.17

Estonia 0.79 0.88 1.19

Finland 0.82 0.85 1.19

France 0.79 0.84 1.24

Germany 0.74 0.93 1.24

Greece 0.9 0.87 1.21

Hungary 0.69 0.91 1.33

Iceland 0.76 0.87 1.27

Ireland 0.93 0.85 1.15

Italy 0.78 1.02 1.26

Latvia 0.75 0.87 1.24



Luxembourg 0.7 0.86 1.26

Netherlands 0.69 0.88 1.3

Norway 0.86 0.89 1.15

Portugal 0.84 0.97 1.3

Romania 1.13 0.84 1.37

Slovak Rep 0.55 0.93 1.27

Spain 0.8 0.83 1.24

Sweden 0.72 0.93 1.15

Switzerland 0.79 0.92 1.15

UK 0.83 0.89 1.18

US 0.7 0.85 1.18

Note: The table reports concentration indexes of job categories by level of education and country. We rely on LFS threefold

classification of education derived from ISCED2011 (low: lower secondary, middle: upper secondary, high: higher education

attainment). For any of the 27 countries of the sample, concentration indexes are computed as the ratio between the share

of jobs of category i in education level j over the share of category i in total employment. Numbers greater (lower) than one

denote over-representation (under-representation) in that specific category. Data refer to 2018.


Table A A. 17. Concentration indexes of category 1 and 3 jobs by nativity status and country

Country Native - IC Cat.1 Native - IC Cat.3 Foreign - IC Cat.1 Foreign - IC Cat.3

Austria 1.09 1.04 0.68 0.84

Belgium 1.04 1.01 0.81 0.94

Croatia 1 1 0.97 1

Cyprus 1.07 1.03 0.78 0.89

Czech Rep 1 1.01 0.89 0.88

Denmark 1.03 1.02 0.79 0.89

Estonia 1.03 1.01 0.73 0.89

Finland 1.02 1.01 0.7 0.84

France 1.03 1.01 0.81 0.91

Germany 1.08 1.03 0.66 0.88

Greece 1.05 1.02 0.4 0.75

Hungary 1 1 1.17 1

Iceland 1.03 1.02 0.68 0.84

Ireland 1.04 1.04 0.86 0.88

Italy 1.1 1.05 0.39 0.71

Latvia 1.02 1.01 0.82 0.92

Luxembourg 0.87 0.91 1.09 1.07

Netherlands 1.02 1.01 0.87 0.94

Norway 1.05 1.03 0.75 0.87

Portugal 0.99 1 1.07 0.97

Romania 1 1 1.92 1.09

Slovak Rep 1 1 1.41 1.15

Spain 1.07 1.04 0.62 0.8

Sweden 1.06 1.04 0.77 0.84

Switzerland 1.06 1.03 0.86 0.94

UK 1.02 1.01 0.92 0.94

US 1.05 1.02 0.81 0.91

Note: The table reports concentration indexes of job categories by nativity status (foreign-born vs. native-born) and country.

For any of the 27 countries of the sample, concentration indexes are computed as the ratio between the share of jobs of

category i in status j over the share of category i in total employment. Numbers greater (lower) than one denote over-representation





Table A A. 18. Concentration indexes of job categories by ISCO 2-digit code

ISCO code

ISCO name IC Cat. 1

IC Cat. 2

IC Cat. 3

11 Chief Executives, Senior Officials and Legislators 2.99 2.16 1.90

12 Administrative and Commercial Managers 2.82 2.12 1.89

13 Production and Specialized Services Managers 1.65 1.50 1.77

14 Hospitality, Retail and Other Services Managers 1.52 1.17 1.18

21 Science and Engineering Professionals 1.81 1.95 1.87

22 Health Professionals 0.00 0.00 0.00

23 Teaching Professionals 1.51 1.08 0.95

24 Business and Administration Professionals 2.87 2.02 1.85

25 Information and Communications Technology Professionals 3.11 2.20 1.90

26 Legal, Social and Cultural Professionals 1.58 1.12 1.53

31 Science and Engineering Associate Professionals 0.39 0.90 0.86

32 Health Associate Professionals 0.14 0.11 0.10

33 Business and Administration Associate Professionals 2.35 1.71 1.64

34 Legal, Social, Cultural and Related Associate Professionals 0.98 0.70 0.67

35 Information and Communications Technicians 2.58 1.96 1.85

41 General and Keyboard Clerks 3.05 2.16 1.86

42 Customer Services Clerks 1.40 0.99 1.24

43 Numerical and Material Recording Clerks 1.51 1.70 1.49

44 Other Clerical Support Workers 0.74 0.53 0.83

51 Personal Services Workers 0.06 0.25 0.33

52 Sales Workers 0.87 0.61 0.54

53 Personal Care Workers 0.01 0.01 0.01

54 Protective Services Workers 0.04 0.03 0.18

61 Market-oriented Skilled Agricultural Workers 0.00 1.50 1.35

62 Market-oriented Skilled Forestry, Fishery and Hunting Workers 0.00 1.45 1.44

63 Subsistence Farmers, Fishers, Hunters and Gatherers 0.00 1.01 1.39

71 Building and Related Trades Workers (excluding Electricians) 0.00 0.25 0.46

72 Metal, Machinery and Related Trades Workers 0.00 1.58 1.42

73 Handicraft and Printing Workers 0.32 1.61 1.53

74 Electrical and Electronic Trades Workers 0.00 0.05 0.20

75 Food Processing, Woodworking, Garment and Other Craft and Related Trades Workers

0.04 0.75 0.92

81 Stationary Plant and Machine Operators 0.00 1.61 1.42

82 Assemblers 0.00 0.60 0.52

83 Drivers and Mobile Plant Operators 0.00 0.40 0.88

91 Cleaners and Helpers 0.00 1.01 0.88

92 Agricultural, Forestry and Fishery Labourers 0.00 0.44 0.67

93 Labourers in Mining, Construction, Manufacturing and Transport 0.00 0.41 0.36

94 Food Preparation Assistants 0.00 0.00 0.06

96 Refuse Workers and Other Elementary Workers 0.04 0.13 0.13

Note: The table reports concentration indexes of job categories by ISCO 2-digit code using data from the 27 countries of the

sample. For any given occupation, concentration indexes are computed as the ratio between the share of jobs of category i

in occupation j over the share of category i in total employment. Numbers greater (lower) than one denote over-representation

(under-representation) in that specific category. ISCO code 95 has been dropped due to inconsistencies between ICP INAPP

and O*NET data. Data refer to 2018.




Table A A. 19. Essential occupations as identified by Fasani and Mazza (2020)

ISCO code ISCO name

213 Life Science Professionals

214 Engineering Professionals (excluding Electrotechnology)

221 Medical Doctors

222 Nursing and Midwifery Professionals

223 Traditional and Complementary Medicine Professionals

224 Paramedical Practitioners

226 Other Health Professionals

231 University and Higher Education Teachers

232 Vocational Education Teachers

233 Secondary Education Teachers

234 Primary School and Early Childhood Teachers

235 Other Teaching Professionals

251 Software and Applications Developers and Analysts

252 Database and Network Professionals

314 Life Science Technicians and Related Associate Professionals

311 Physical and Engineering Science Technicians

312 Mining, Manufacturing and Construction Supervisors

313 Process Control Technicians

315 Ship and Aircraft Controllers and Technicians

321 Medical and Pharmaceutical Technicians

322 Nursing and Midwifery Associate Professionals

351 Information and Communications Technology Operations and User Support Technicians

352 Telecommunications and Broadcasting Technicians

511 Travel Attendants, Conductors and Guides

516 Other Personal Services Workers

531 Child Care Workers and Teachers’ Aides

532 Personal Care Workers in Health Services

612 Animal Producers

613 Mixed Crop and Animal Producers

611 Market Gardeners and Crop Growers

751 Food Processing and Related Trades Workers

816 Food and Related Products Machine Operators

831 Locomotive Engine Drivers and Related Workers

832 Car, Van and Motorcycle Drivers

833 Heavy Truck and Bus Drivers

835 Ships’ Deck Crews and Related Workers

911 Domestic, Hotel and Office Cleaners and Helpers

912 Vehicle, Window, Laundry and Other Hand Cleaning Workers

933 Transport and Storage Labourers

961 Refuse Workers

Note: The table lists the ISCO 3-digit occupations identified as “key” in the work by Fasani and Mazza (2020), i.e. occupations

that need to be performed even during a pandemic in order to keep citizens healthy, safe and fed.

Source: Fasani and Mazza (2020).

Table A A. 20. Essential workers by job category in European countries and in the US

Country Essential workers

Essential workers (% of

total)

Cat. 1

(% of essential)

Cat. 2

(% of essential)

Cat. 3 (% of essential)

Unsafe

(% of essential)

Austria 1,503,010 0.35 0.21 0.42 0.47 0.53

Belgium 1,790,841 0.38 0.19 0.34 0.39 0.61

Croatia 543,614 0.33 0.16 0.39 0.45 0.55

Cyprus 120,257 0.31 0.19 0.38 0.43 0.57



Czech Republic 1,557,736 0.29 0.22 0.39 0.47 0.53

Denmark 1,222,528 0.43 0.20 0.35 0.38 0.62

Estonia 195,221 0.30 0.24 0.37 0.45 0.55

Finland 1,009,995 0.40 0.22 0.37 0.42 0.58

France 10,978,456 0.41 0.15 0.34 0.39 0.61

Germany 12,956,722 0.31 0.21 0.38 0.43 0.57

Greece 1,433,295 0.38 0.15 0.41 0.46 0.54

Hungary 1,326,309 0.30 0.17 0.37 0.46 0.54

Iceland 69,299 0.35 0.21 0.32 0.36 0.64

Ireland 727,134 0.33 0.16 0.37 0.41 0.59

Italy 7,494,234 0.33 0.18 0.36 0.42 0.58

Latvia 272,169 0.30 0.13 0.30 0.39 0.61

Luxembourg 85,329 0.32 0.21 0.39 0.45 0.55

Netherlands 2,989,351 0.35 0.22 0.36 0.40 0.60

Norway 1,082,006 0.41 0.18 0.31 0.34 0.66

Portugal 1,539,085 0.32 0.20 0.37 0.43 0.57

Romania 3,559,995 0.41 0.11 0.59 0.65 0.35

Slovak Rep. 760,054 0.30 0.15 0.27 0.35 0.65

Spain 6,766,205 0.36 0.17 0.34 0.39 0.61

Sweden 2,059,073 0.40 0.21 0.34 0.38 0.62

Switzerland 1,483,723 0.34 0.23 0.41 0.45 0.55

UK 10,616,055 0.33 0.22 0.34 0.38 0.62

US 45,366,814 0.29 0.20 0.31 0.35 0.65

Note: The table reports details on the number and the distribution across the four job categories of our taxonomy of essential workers, i.e.

individuals holding a job in any of the occupations defined as “key” by Fasani and Mazza (2020).




Annex B. Additional Figures

Figure A B. 1 Concentration indexes of category 3 jobs by age group and country

Note: The figure above shows concentration indexes for job category 3 by age group across the 27 countries of the sample. Concentration

indexes are computed as the ratio between the share of jobs of category 3 for group j over the share of category 3 in total employment. Numbers

greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Data refer to 2018.




Figure A B. 2. Concentration indexes of category 1 jobs by age group and country

Note: The figure above shows concentration indexes for job category 1 by age group across the 27 countries of the sample. Concentration

indexes are computed as the ratio between the share of jobs of category 1 for group j over the share of category 1 in total employment. Numbers

greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Data refer to 2018.




Figure A B. 3. Concentration indexes of category 3 jobs by gender and country

Note: The figure shows concentration indexes for job category 3 by gender across the 27 countries of the sample. Concentration indexes are

computed as the ratio between the share of jobs of category 3 for gender j over the share of category 3 in total employment. Numbers greater

(lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Data refer to 2018.




Figure A B. 4. Concentration indexes of job categories by type of job

Note: The figure shows concentration indexes for job categories by type of job. Concentration indexes are computed as the ratio between the

share of jobs of category i for type j over the share of category i in total employment, pooling data from the 27 countries of the sample. Numbers

greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific category. Vertical bars

measure one standard deviation above and below the cross-country average of concentration indexes. Data refer to 2018.




Figure A B. 5. Age distribution of vulnerable and total workers

Note: The figure shows the age distribution for total workers and vulnerable workers (low educated individuals with an unsafe job in a firm with

less than 20 employees in the “Accommodation and food service activities” or “Arts, entertainment and recreation” sectors). Shares are computed

as the number of vulnerable (total) workers in a given age bracket over the total number of vulnerable (total) workers, pooling data from 25

countries (data are not available for Latvia and the United States). Data refers to 2018.




Figure A B. 6. Age distribution of unsafe non-essential and total workers

Note: The figure shows the age distribution for total workers and unsafe non-essential workers (as defined in Fasani and Mazza, 2020). Shares

are computed as the number of unsafe non-essential (total) workers in a given age bracket over the total number of unsafe non-essential (total)

workers, pooling data from 25 countries (data are not available for Latvia and the United States). Data refers to 2018.




Figure A B. 7. Distribution of unsafe non-essential workers over economic sectors

Note: The figure shows the distribution for unsafe non-essential workers over economic sectors (Nace rev 2). Shares are computed as the

number of unsafe non-essential (total) workers in a given sector over the total number of unsafe non-essential (total) workers, pooling data from

26 countries (data are not available for the United States). Data refers to 2018.




Figure A B. 8. Concentration indexes of job categories by ethnicity

Note: The figure shows concentration indexes of job categories by ethnicity. Concentration indexes are computed as the ratio between the share

of jobs of category i for ethnic group j over the share of category i in total employment. Data refer to US in 2018 (data are not available for EU

countries). Numbers greater (lower) than one (horizontal dashed bar) denote over-representation (under-representation) in that specific

category.

Source: Current Population Survey (CPS).

The New Hazardous Jobs and Worker Reallocation - OECD

Documents